Dan Margulis Applied Color Theory

Color Theory and Practice

Color theory and practice
Posted by: Boris Feldblyum
Fri Apr 27, 2007 1:15 am (PST)

Gentlemen,

Whether the list will move or not, the name of the game for me remains the same: how to create high impact images.

Here is one: [ http: //www.bfcollection.net/errordocs/02.htm
]http://www.bfcollection.net/errordocs/02.htm

I spent way too much time on it today and am no longer sure I did the best job possible. Perhaps it will be clearer in the morning.

This is not about the lightest or darkest "significant" parts of the image. For the client (architect) everything is significant. It is a photo taken at dusk, which is always beautiful and challenging time of the day. I am not sure right now if it can be done better; it's just the lingering feeling: the picture lack the final magic touch that will make it jump out of paper.

Am I right?

Any and all commnts will be greatly appreciated.

Boris Feldblyum
www.bfcollection.net
___________________________________________________________________________

Re: Color theory and practice
Posted by: Stephen Marsh
Fri Apr 27, 2007 9:02 am (PST)

Boris Feldblyum wrote:

I spent way too much time on it today and am no longer sure I did
the best job possible. Perhaps it will be clearer in the morning.

I would be interested in seeing the original at the same resolution as the worked example Boris.

As for your edits, I know what you mean, it can be hard to milk such types of images and it is very open to artistic interpretation. Forgetting the sky, the area of interest is low key. One wishes to engineer more contrast into the area of interest, while still keeping the impression of the original lighting of the scene (for both artistic and technical reasons related to print).

I am not sure right now if it can be done better; it's just the
lingering feeling: the picture lack the final magic touch that will
make it jump out of paper.

With a quick glance, I would say that it is mostly there Boris, but more may be needed for impact in print.

Not having access to the original, I would recommend use of the shadow/highlight command and then further contrast enhancing moves (to counter the previous edit). Back in the day I did this using contrast masking or similar manual methods, combined with reduced opacity "light" layer blending modes (CMYK blends different to RGB and one may prefer the CMYK edit, even if the final destination is RGB) that were masked to exclude the transitions into the endpoints - and curves too of course (steep curve in this area of interest)! Sometimes one may multiply blend in the K channel of a UCR sep at reduced opacity with layer blend if to further alter the lighting.

Then there is ACR4 and Photoshop Lightroom (highlight recovery, fill light, tonal edits, exposure, vibrance etc).

Various sharpening methods could be the next point (previous tonal edits will have a contrast/sharpening effect anyway).

Regards,

Stephen Marsh.
___________________________________________________________________________

Re: Color theory and practice
Posted by: "dandavide0"
Fri Apr 27, 2007 10:30 am (PST)

Hi Boris

This is my first time writing something on this group and i'm not a professional but simply a very very passionate about color correction and photography.

Maybe you are right this picture missed the final touch. I've just finished to read the last Dan's book. It was very hard for me both for the language, because I'm Italian and I've learned English reading this book, and for the contents, but nevertheless let's us continue with the picture. Following some Dan's techniques I tried to add two new layer as a copy of the background on RGB space. In the middle layer I applied the inverted red channel and then gausian blur with a radius of 70, layer mode set up on overlay. The top layer mode, which is a copy of the background, set up on darkness then the opacity on 50%. In order to leave untouched the sky I added to this top layer an additional layer mask applying on it an inverted copy of the blue channel. If you like you can also make some adjustment on the layer mask just created with curve which will allow you to fine tune the power of the layer mask.

This is my idea and I will appreciate comments If I'm in the right way or in the wrong one Sorry for my English language I hope in your understanding

Davide D'Angelo

Italy
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Fri Apr 27, 2007 10:41 am (PST)

Boris,

I think this is a stunning photograph. I would be interested to learn about how you so carefully controlled the highlight and shadow areas, because these are very much at risk in such lighting conditions. The shadow detail shows beautifully on the monitor; however, I'd be curious to know what media you would use for printing this photograph, and whether in a print the foreground would need to be very gently lightened with a slight increase of local contrast to maintain that kind of detail on paper.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: "chris broadhurst"
Fri Apr 27, 2007 10:43 am (PST)

Hi, lovely image Boris, I would lighten the odd bits, like...

http: //www.broadhurst-family.co.uk/Photos/20070413023.jpg

I leave you to sharpen etc.

Regards
Chris
http://www.broadhurst-family.co.uk/lefteye/
___________________________________________________________________________

Re: Color theory and practice
Posted by: Ric Cohn
Fri Apr 27, 2007 10:58 am (PST)

Boris,

Really nice image. I do think it can be made better, but any changes from where you are would IMHO be a matter of opinion, not right or wrong. What color space is it suppose to be in? I actually prefer it in ColorMatchRGB which might be lighter than you mean it to be.

After assigning ColorMatchRGB I think the sky could be darker and the silver doors lighter and the brightness of the reds brought down so as not to draw too much attention. I did a quick adjustment that I like better using Apply with Blend-If and Curves in RGB and Lab, followed by a healthy dose of Hiraloam sharpening . A larger file to look at would be nice.

Ric Cohn
___________________________________________________________________________

Re: Color theory and practice
Posted by: J Walton
Fri Apr 27, 2007 2:46 pm (PST)

On 4/26/07, Boris Feldblyum wrote:

Any and all commnts will be greatly appreciated.

We'll see about that! :-)

My main comment has to do with the crop. It seems like you worked really hard on the middle half of the image but there's just not much interest in the bottom. I can see that there's not much you can do with the sky, but I'm confused by the area at the bottom.

It either needs to be more attractive or needs to be cropped out. Right now it's like this big, dark, bluhhhhh at the bottom of a striking image. You could darken it down and crop it out, or cut some quick masks and brush in some lighting to create interest. I'd start by brushing in a light yellow on a layer set to overlay, and see how that looks. Remember the brush should be big and the opacity small.

--
J Walton
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Lee Clawson"
Fri Apr 27, 2007 2:49 pm (PST)

Boris,

I agree with your "lingering feeling". The exterior lighting is verging on flat. For an architectural shot I want it to be a bit more dramatic.

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Adriano Esteves"
Fri Apr 27, 2007 2:53 pm (PST)

Boris,

Very nice photo indeed!. I would remove the electric blue from it (my take was to blend 30% of the the RED in darken mode), after that I went to LAB and reduce the Magenta on the wall (maybe cMyk would be better). I gained a greener look on the bottom right construction but it doesn't seems that wrong. The sky now seem more natural to me and the rest of the colors are still there.

here's my take: http: //photo.adrianoesteves.com/color/try.jpg

Hope you can give us a higher res :)

later,
Adriano Esteves
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Alessandro Bernardi"
Sat Apr 28, 2007 3:27 am (PST)

Boris wrote;

This is not about the lightest or darkest "significant" parts of the
image. For the client (architect) everything is significant. It is a photo
taken at dusk, which is always beautiful and challenging time of the day.
I am not sure right now if it can be done better; it's just the lingering
feeling: the picture lack the final magic touch that will make it
jump out of paper.


Yes I agree with that it's not about the lightest or darkest "significant" parts of the image but here you have a problem of "too much darkness", due to the sky that for me it's very distracting because it's too luminous. I think that the architect will appreciate more to see better the building than the sky....

So even if it's a particular moment of the day, the key is to darken the sky with a more dramatic color and the lower part will take more life...

I also see too much magenta on the file (I don't konw if is a problem of profiles of something else) and I would fix this.

I tried some moves in LAB but the problem is that when you convert the file in CMYK the final result is not so pretty, and I think you'll have to give a CMYK file to the printer or not?

In any case you can use an RGB version making it from the CMYK file if you like it.

So I made this:

- duplicate and convert the image in CMYK SWOP Coated 20%, GCR Medium 95-300-0

- duplicate the Black channel, invert, copy and paste onto the RGB file in Overlay mode

- Gaussian Blur of 92 pixels: this gives you the light in the lower part of the image and darkens the sky

- adjust the amount of contrast obtained by applying a curve at the layer for making the darkest gray level of about 53 RGB value and leaving the lightest gray level untouched

- flatten and convert in CMYK with the same settings above

- apply an adjustment layer->Selective Color to make the Reds more warm with some extra Yellow, eliminate some Green casts adding Magenta and subtracting Cyan and Yellow in the Greens, subtract Magenta from the Blues 25% and add extra Cyan 35%

- apply an Adjustment layer->Curves with an S shape in the Black channel to have more contrast and light also making the shadows deeper.

This is for me enough for making this image more interesting.

If you prefer you can do some sharpening in the black channel.

Here are my RGB and CMYK PSD files: http: //www.bbexecutive.com/boris.html

I hope you'll like them.

Alessandro Bernardi
___________________________________________________________________________

R: [colortheory] Color theory and practice
Posted by: "Davide D'Angelo"
Fri Apr 27, 2007 2:48 pm (PST)

Hi Boris

This is my first time writing something on this group.

Maybe you are right this picture missed the final touch. I've just finished to read the last Dan's book. It was very hard for me both for the language, because I'm Italian and I've learned English reading this book, and for the contents, but nevertheless let's us continue with the picture. Following some Dan's techniques I tried to add two new layer as a copy of the background on RGB space. In the middle layer I applied the inverted red channel and then gausian blur with a radius of 70, layer mode set up on overlay. The top layer mode, which is a copy of the background, set up on darkness then the opacity on 50%. In order to leave untouched the sky I added to this top layer an additional layer mask applying on it an inverted copy of the blue channel. If you like you can also make some adjustment on the layer mask just created with curve which will allow you to fine tune the power of the layer mask.

This is my idea and I will appreciate comments If I'm in the right way or in the wrong one Sorry for my English language I hope in your understanding

Davide D'Angelo

Italy
___________________________________________________________________________

Re: Color theory and practice
Posted by: R. Lutz
Sat Apr 28, 2007 6:07 pm (PST)

It has a lot of color and looks pretty nice as is, Boris. I agree with some of the other posters that a darker sky and a lighter pavement would be a help. If you can get away with it, you might stretch the image. The effect is similar to using a longer lens on the camera. This seems to make the image a little more inviting.

I posted an example in the file section for this list. The name is "store.jpg" . I'll remove it in a few days. Go to http://tech.groups.yahoo.com/group/colortheory/files/

Dick Lutz
___________________________________________________________________________

Re: Color theory and practice
Posted by: "scotty8082001"
Sat Apr 28, 2007 6:08 pm (PST)

O.K. ... here I go off the deep end. This is my first post to this group, and with all the heavyweight pros here, I'll probably get pilloried as a color bumpkin, but nonetheless ...

In PP5, Dan talks about "betting the image." My approach to this shot is to "bet the image" that the central vertical slab is really black (if not, then all bets are off). If it is black, then at the top it is reflecting the sky, and since black is neutral, the proportions of R/G/B in the reflected sky should be the same as the proportions of R/G/B in the actual sky, so ...

First, I added a Levels adjustment level and clicked the black point eyedropper in the darkest part of the concrete basin at the bottom of the picture (sampler point #3; see the screen cap <http: //www.simpsonsresearch.com/Boris/Boris.html>that shows where I put my sampler points). This put a bit more tooth in the foreground.

Second, I sampled the darkest part of the sky (sampler point #1) and the darkest part of the reflected sky in the central vertical slab (sampler point #2). Adding a Curves adjustment layer, I adjusted the Green and Blue channels at sampler point #2 so that the ratio of R/G and the ratio of R/B was the same as in the sky at sampler point #1. (Since Red is a major color in the image, and since I wanted to keep the Red the way it is, that's why I used the Red channel as my ratio base, adjusting the Green and Blue.)

You can see the side-by-side results here:
http://www.simpsonsresearch.com/Boris/Boris.html
<http: //www.simpsonsresearch.com/Boris/Boris.html> .

To my eye, this does a couple of things:

1. It neutralizes the left wall of the building, which I see as gray concrete, and

2. It darkens the foreground which moves the eye up into the center-of-interest. It seems to increase the "depth" of the image and makes it "pop" a bit more.

3. I like the sky better being a bit more cyan.

So, do I win the image-bet, or lose it?

Best regards,

Scott Simpson
Simpson's Research
Fort Wayne, IN
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Murray DeJager"
Sat Apr 28, 2007 6:10 pm (PST)

Hi Boris,

This is my take on your photo.

I'm not at all proficient with blending channels and stuff like that yet so I had to use the tools I know a little about.

First of all, however, I thought the only thing really wrong with your version of the photo was that the foreground was too dark and flat. I don't think it was really that dark out when you took the photo and I don't think someone standing where you were when you took the photo would have seen the foreground to be as dark as you made it.

With that in mind the first thing I did was to lighten the darkest tones with the S/H command.

That done I converted to LAB and gave the "L" channel an "S" curve. This gave the photo a little "snap!" Or is that "punch!" I don't remember from Dan's books.

From their I converted the photo to sRGB. Actually, that brings up a point. Your photo wasn't tagged but in my opinion it looked best with the AdobeRGB profile assigned to it at the beginning of all this work.

From there the only thing I did was to make the sky a little darker with the Replace Color command and a layer mask to keep the changes confined to the sky. And that's just a personal opinion, I think your sky was fine.

I added the photos to the Photos section in an album called Boris' Architecture. (I don't have my own web site.) I hope you don't mind Boris. Maybe Dgrin would be a much better place to discuss and share these kinds of discussions!

Murray DeJager
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Sun Apr 29, 2007 1:11 am (PST)

Scott,

This was courageous and honest on your part; and I respect what you were trying to do - stating the problems as you saw them and deploying the tools you know for resolving them; this is straightforward straight-shooting. However in the final analysis, looking at the result on my monitor, I think Boris's original is better. I found your outcome a bit more dense than his, especially in the places where one wouldn't want that (for example, bottom foreground, lower parts of the building). But it was good that you posted the results of your effort, allowing us to see the outcome of the moves you described. Please don't let a comment like this discourage you from continuing to contribute your ideas and thinking. Maybe others will disagree with my evaluation of this effort. As well, we've seen from other posted results that (at least in my opinion) they didn't succeed in improving anything either, hence it is clear that fiddling with this image is not an undertaking for the feint-hearted.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: Richard Wagner
Sun Apr 29, 2007 1:15 am (PST)

Boris,

Here is another interpretation of your image with corrections that would be appropriate for this image to go to print, with the intention of showing off the building for the architect. The corrections were made in about 15 minutes max. sRGB was assumed as a starting point, and the corrections were made in five layers using standard techniques. The .psd file is located in the Files section as 20070413023_mod.zip
(http://tinyurl.com/3x8bly) and the jpg can be found here (http://tinyurl.com/3xl3tk).

Best,

--Rich Wagner
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Bruce Albrecht"
Sun Apr 29, 2007 4:37 am (PST)

Boris,

The subjective nature of the colors in the original scene make it quite difficult to decide what is the "correct" way to push this excellent image. The cool blue of the open sky, the yellow bath of the light cans on red brick, the orange-red light of the interior...I would suggest the only person who can target the original colors is the individual who was behind the lens.

For my part, posted at http: //www.punapix.com/tests/boris042907.jpg, I found the original--and many of the corrections submitted so far--to be too blue, particularly in the concrete. I also agree with others that the foreground is too muddy. My correction was to convert to LAB, apply a curve adjustment layer to neutralize the blue in the concrete, and lighten the quarter-tone. I limited the effect by using an inverted blue channel for the mask. I also found it aesthetically pleasing to steepen the a channel by 5 points. Did I go too far? I think the extra red counters the cool blues well, but is it accurate? Or OOG for the output method? Most importantly, will the client appreciate it?

I also find the composition more pleasing if cropped a little tighter, if the format will allow for it.

These are the types of posts where this list really shines, IMHO. While I have learned an awful lot from the endless arguments over such affairs as proper profiling, it is the collective input of so many talented professionals on a particular image that I really enjoy. If dgrin makes this type of thread easier to digest then it's worth the extra hassle.

Bruce Albrecht
___________________________________________________________________________

Re: Color theory and practice
Posted by: Richard Wagner
Sun Apr 29, 2007 6:17 pm (PST)

So much for Yahoo Groups and images. The files I posted are still in the Files section of Yahoo Groups - the links that were posted previously are broken. The files can also be accessed here, as well as through Yahoo Groups:

ftp: //www.wildnaturephotos.com/download/20070413023_mod.jpg

Layered .psd file.
ftp: //www.wildnaturephotos.com/download/20070413023_mod.zip

--Rich Wagner
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Ben Richardson"
Sun Apr 29, 2007 6:18 pm (PST)

Hi all, this is my first post. Though I've been reading for some time, it's just now I finally felt I had something to add!

My version is at http://benrichardson.com/colortheory/ It's been corrected using a kind of digital ENR, or bleach bypass: a photochemical technique used for motion picture release prints. This gist of it is that you end up with a black and white image overlaid on the colour image, which enhances midtone contrast, adds density and desaturates the image.

How I get there varies by image, but it this case it began with a Hue/Sat adjustment layer at -100% saturation, set to Multiply. This, of course, makes the image way too dark, but only through the mids, so a strong correction either with Levels' mid-slider or an aggressive curve is all that's required to bring it back (I used a Curves layer with points at 19/66 and 91/198 set to Luminosity). I grouped these two layers and adjusted opacity by eye (50%, in this case). I also added a Channel Mixer layer (monochrome from 100% green) at 35% Soft Light and a second Hue/Sat at +30% saturation, set to Darken. Finally, judging the sky still too light, I added a darkening Curve at the bottom of the stack, using the Blend If sliders to limit its effect to the sky.

I'd be happy to hear everyone's thoughts.

Ben Richardson
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Sun Apr 29, 2007 10:23 pm (PST)

I don't think it would print as well as the image posted by Rich Wagner; with a series of rather ordinary Curves Adjustment Layers and gradients that image picks up the low-lights and clarifies areas of under-exposure enough to produce a vibrant, detailed print. From what I see on my monitor I don't think your image would do this as effectively.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Mon Apr 30, 2007 3:35 pm (PST)

This picture is an extreme example of how humans and cameras evaluate scenes differently. We all know how offensive cross-lighting can be. Here, because of all the different light sources, the top left side of the building has a bluish-purple cast, but the cement in the foreground is greenish-blue. If we were at this scene we wouldn't necessarily see them as the same hue but they would be very much closer than what the camera picked up. Those corrected images that I've looked at from list members all are much too blue or purple in the background.

Also, there are two subareas that have the same type of problem. The strong yellow lighting in the first-story rooms overwhelms the camera--we would not see those areas as nearly as yellow. And on the second floor the camera doesn't neutralize the light shining through the blue windows nearly as well as our eyes would. When this effect occurs we typically have to use the B channel of LAB as a mask for the area and then reduce its saturation.

I threw a quick version into the Filess section of the list, labeled Boris/Dan. Obviously many different views are possible on how dark everything should be, as well as how to treat the yellow and blue areas. But there shouldn't be much dispute that we need to move toward neutrality in both the foreground and background cement. Doing this while avoiding a selection requires careful curving. In images like this, even though they are headed for RGB, it would be well to keep the possibilities of CMYK in mind: the presence of the black channel can help maintain neutrality in the desired areas while not muddying up the sky, which wouldn't contain any black.

I actually used two sets of curves: one in RGB to get somewhat close, reverting to Color mode to preserve detail, and then one in CMYK. I went to LAB afterward to try to get more color into the foreground brickwork.

If anybody who did a version doubts the necessity of getting these neutrals closer, paste a copy of mine on top of yours, setting layer mode to Color. That will eliminate any difference of opinion as to how dark the image should be and let you see the impact of the color shift only.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Michael Demyan"
Mon Apr 30, 2007 7:23 pm (PST)

Hi Group:

There are all kinds of contortions we can go through to make the image pleasing.

Without being there and viewing the scene when it was photographed, correcting it will be by using the tools that work best for us, and making the image look the way we might see it in our mind's eye.

So for want it is worth - here is my mind's eye view of how the scene might have looked. Working in AdobeRGB.
http: //www.mikedemyan.com/dpreview/ColorTheory%20and%20My%20Eye.jpg

All I used to correct the image is one commercial plug-in "Shadow Illuminator" to bring out the shadow detail, and the balance of corrections were done using Hue/Saturation.

Mike
Fine Photography by
Michael Demyan
www.mikedemyan.com
610-758-9769
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Murray DeJager"
Mon Apr 30, 2007 7:24 pm (PST)

Hi Dan,

Yup, I agree. To be fare to myself I didn't even think of pulling out the Color Sampler Tool, I thought I'd give Boris the benefit of the doubt and assume that the final colors were fine with him. I also, however, never thought the colors were that far out though. When I saw Richard's version I realized my foreground cement was to green so I made a slight change to that, but that was it.

I guess thats why you make the Big Bucks!

Murray DeJager

If anybody who did a version doubts the necessity of getting these
neutrals closer, paste a copy of mine on top of yours, setting
layer mode to Color. That will eliminate any difference of opinion
as to how dark the image should be and let you see the impact of
the color shift only.
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Tue May 1, 2007 8:57 am (PST)

Michael,

I think your first couple of sentences are exactly right. In fact, it is a bit hazardous going too far on this image without a more comprehensive statement of intent from the client and the photographer. But one can identify and correct things that could interfere with achieving the most readily supposed objectives of the image and this you and the image from Rich Wagner have done. Along with the image that Rich Wagner posted I think this is the most successful re-take of all that have been posted - and for the same reason - the atmosphere is preserved while lifting the heavier areas that would give the most difficulty printing. This is a bit of a tricky image to work-over because the sky and the atmospherics on the building get wrecked if one gets too fancy or too heavy-handed with it. You mention using the "Shadow Illuminator" plug-in. I am interested in learning about it. Could you point me to the vendor's website, and at the same time I'd appreciate your take on how it differs from Photoshop's Shadow/Highlight filter.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: André Dumas
Tue May 1, 2007 8:58 am (PST)

Hello Boris

Your client is an architect. If this is one picture out of many to describe that building then, in my opinion, it is a good picture. At dusk, few things are neutral, artistically speaking it is not too blue and it is not too dark, it is what you have made it. On the other hand if this is the only picture to describe the entire building then, from an architect's point-of-view, it is not a photo that describe the building adequately.

You say "I am not sure right now if it can be done better". It depends what "better" means to you, every one has a personal opinion on that and you have seen quite a few so far.

In my opinion a series of black and white photos best describe a building, its design and characteristics, architecturally speaking that is.

André Dumas
___________________________________________________________________________

Re: Color theory and practice
Posted by: "scotty8082001"
Tue May 1, 2007 10:30 am (PST)

Thanks for posting this image, Dan. Just looking at it with color sampler tool in hand taught me a lot. Neutralizing the concrete was what I was shooting for in my earlier effort, but your example showed me how much I missed the mark. That's why yo' da man ...

Best regards,
Scott Simpson
Fort Wayne, IN
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Lee Clawson"
Tue May 1, 2007 8:58 pm (PST)

on 4/30/07 6:28 PM, Dan Margulis wrote:

If anybody who did a version doubts the necessity of getting these neutrals
closer, paste a copy of mine on top of yours, setting layer mode to Color.
That will eliminate any difference of opinion as to how dark the image should
be and let you see the impact of the color shift only.

Dan,

We have a fairly good idea that this is to highlight an architect's design as described in the OP below:

This is not about the lightest or darkest "significant" parts of the image.
For the client (architect) everything is significant. It is a photo taken at
dusk, which is always beautiful and challenging time of the day. I am not
sure right now if it can be done better; it's just the lingering feeling: the
picture lack the final magic touch that will make it jump out of paper.

I keep wanting to minimize the distracting changes of shape, color and light that make up this image of the building.

A rear block wall and the concrete shape in the foreground that's (now) neutral and a walkway with more color, both now lighter, (to my eyes) also adds competing visual elements. For me they are minor concerns versus bringing visual focus to a photograph of a building.

I'll be very curious to hear what the client says...

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Paul Marriner"
Tue May 1, 2007 8:59 pm (PST)

Mark, I'm happy you asked. I was going to write earlier about SI but wasn't sure it would get by Stephen. I've used this plug ever since Vlad introduced it. I also use S/H and Digital SHO Pro as well as several actions and the tools in LR and ACR. Why? Because I deal with thousands of tranny scans from a CanoScan 4000. This scanner, like so many digital sensors, has mediocre DR. I invariably choose to hold the highlights, which results in plugged shadows. Ergo, I need an array of tools to recover shadow detail.

In many instances, SI does a better job than the others, including S/H, with far less work (although on images larger than 50MB it's a bit slow). The defaults are frequently bang on. Take it for a trial run. I've included an interesting note on its history.

www.shadowilluminator.org/

Paul Marriner

Vladimir Brajovic is a robotics researcher at Carnegie Mellon University. His goal is to give a better vision sense to robots. But in an unexpected side-effect, he developed a smart system to automatically enhance underexposed photos. The result, the Shadow Illuminator, is quite convincing.

The Shadow Illuminator, funded through a $350,000 grant from the National Science Foundation, was developed originally to help robots see better. Using principles based on the physics of how optical images are formed, Shadow Illuminator imitates the vision processes that take place in the human eye. It examines the content of a photograph, estimates the illumination conditions and then brightens shadows. It also enhances details within the shadow.

"Shadow Illuminator is intelligent and works consistently for all pictures," said Brajovic, director of the Computational Sensor Laboratory in Carnegie Mellon's Robotics Institute. "It provides the same results quickly and eliminates the hassle of manually adjusting photographs."
--
Paul Marriner
Outdoor Writing & Photography. Owner: Gale's End Press. Member: OWAA &
OWC. Author of: A Compendium of Canadian Fly Patterns (co-author),
Stillwater Fly Fishing: Tools & Tactics, How to Choose & Use Fly-tying
Thread, Modern Atlantic Salmon Flies, Miramichi River Journal, Ausable
River Journal, and Atlantic Salmon.
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Michael Demyan"
Tue May 1, 2007 8:59 pm (PST)

Hi Mark:

Here is the link to their web site.
https://www.intrigueplugins.com/

Excerpt from the web site:
===============================
Shadow IlluminatorT Pro and Shadow IlluminatorT Home are image processing plug-ins that intelligently restore details in shadows. Through physical modeling of light and surfaces, Shadow IlluminatorT technology compensates for common illumination problems in photographs. Shadow IlluminatorT technology has been originally developed at the Robotics Institute at Carnegie Mellon University to enable robots to approach the visual capabilities of the human eye. Users will find that the results of Shadow Illuminator resemble what their eyes would see if they viewed the environment instead of a camera.

Shadow IlluminatorT plug-in is compatible with most popular image manipulation programs such as Adobe Photoshop, Adobe Elements, Adobe Illustrator, Jasc Paint Shop, Microsoft Digital Image Pro, Irfan View, and others.

===========================================

It is very inexpensive: $30 for a home license and $40 for the pro version.
You can download a trial version also.

See what lurks in the shadows! Have fun, don't make work, correcting images.

Mike


Fine Photography by
Michael Demyan
www.mikedemyan.com
610-758-9769
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Wed May 2, 2007 7:24 am (PST)

Michael,

Thank you, I found it and corresponded with them. They say they have thoroughly tested it for Photoshop CS, they've heard it works with Photoshop CS2, it is not compatible with Photoshop CS3 and they say it will be some months before they produce a version that is. That's OK because there are some critical teething problems with CS3 that make CS3 itself incompatible with me for the time being.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Wed May 2, 2007 7:28 am (PST)

Paul, thanks for the recommendation and explanation. Perhaps a bit OT and you won't be offended if I suggest that much time doing such adjustments can be saved (regardless of how slick the corrective software is) if one improves the "capture conditions". For my legacy film stuff I'm using a Minolta DiMage Scan-Elite 5400, which has excellent rendition of detail and DR, and combined with Silverfast leaves not much to be done in Photoshop for print. [I've written material about this on Luminous-Landscape, and developed material in support of Harald Johnson's "Mastering Digital Printing Second Edition", pages 83-89.] Unfortunately Minolta vacated this business, but Nikon has several comparable models that work faster.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: Stephen Marsh
Wed May 2, 2007 7:28 am (PST)

Re: Color theory and practice

Paul Marriner wrote:

Mark, I'm happy you asked. I was going to write earlier about SI but
wasn't sure it would get by Stephen.

Paul, I am at a loss as to why you would think that talk of "Shadow Illuminator" would not be approved, in the context of making this image "better". My moderation message was about ending potentially endless talk about missing profiles and not talk of image editing, I was wanting talk of image editing to outweigh the missing profile. "SI" would come under the subject of image editing and even if not Photoshop, it is worthy of some minor discussion (just as other third party sharpening or extraction or colour plugs get plugged).

After all, I did write about "SI" in 2003, to name one instance:
http: //tech.groups.yahoo.com/group/colortheory/message/6833

BTW, on some newer cameras from Nikon - there is now an in-camera option for processing JPEG's - called "D-Lighting". It is also presented as part of a stand alone PC/Mac application named "PictureProject". An example has been uploaded to the files section of this website, named Nikon-D-Light-Boris.jpg (run at low setting, there is also normal and high options with no sliders or interface, it is a simple drop down menu choice).

Regards,

Stephen Marsh.
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Howard Smith"
Wed May 2, 2007 8:58 pm (PST)

Paul,

It's something worth trying all right. If you want the same effect in CS3, presuming that CS3 doesn't differ all that much from CS, try creating a Curves Adjustment Layer set to Screen mode. Use 100% opacity. Flatten the image and use Shadow/Highlight to adjust both shadows and highlights. Perhaps not as quick as SI, but the results are essentially the same when using their example.

Howard Smith
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Howard Smith"
Wed May 2, 2007 8:58 pm (PST) Lee,

We seem to think alike on this one. My first impulse was to get in there and start changing things around. Then I realized that Boris had shot this image at dusk for a reason-to add drama and mood to the building. Apparently it had been a clear day, so there would have not been any observable reason not to do the shoot earlier to produce an architecturally correct image with all the correct details and colors. My only suggestion to Boris was that he might want to do a very slight enhancement of the golds and reds to draw more attention directly to the building. For this image it would appear that the surroundings were not considered relevant. The colors in the image would appear to have been consistent with the lighting at that time of day. Photographers have often commented that colors change in an image shot in late afternoon light. When we start changing the colors, we lose some of the effects of the natural lighting. It makes a more acceptable image, but you lose some of the flavor.

From a realistic standpoint, our opinions don't really matter. I, too, am curious to hear what both Boris and his client think of the image. We probably won't agree, but the client pays the bills.

Howard Smith
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Thu May 3, 2007 7:01 am (PST)

Howard,

Once you flatten an image and proceed to build a new structure you are engaged in "destructive editing", which is generally not a preferred approach - one wants to edit while preserving as much reversibility as possible. So I took your basic idea below one step further. Instead of flattening the image. Select but don't open the highest layer in the stack. Then implement a CTRL+ALT+SHIFT+E (Windows - I suppose with Mac it would be CMD+OPTION+SHIFT E) to create a stamped layer (which is everything below it sandwiched into one. Next, your stuff: create a Curves Adjustment Layer in Screen mode clipped to the Stamped layer. This Curve itself can be used for further work on image luminosity in Screen mode. But if that isn't good enough, then select the Stamped Layer, open Shadow/HIghlight filter and go to work on it. You will see by switching it on and off that all these adjustments are confined to the Stamped Layer, so if in the final analysis you don't like the result, you can trash that layer and everything is back to where it was before you started.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Thu May 3, 2007 7:01 am (PST) Murray writes,

Yup, I agree. To be fare to myself I didn't even think of pulling out
the Color Sampler Tool, I thought I'd give Boris the benefit of the
doubt and assume that the final colors were fine with him. I also,
however, never thought the colors were that far out though...I guess thats why
you make the Big Bucks!

and Scott adds,

Thanks for posting this image, Dan. Just looking at it with color
sampler tool in hand taught me a lot. Neutralizing the concrete was
what I was shooting for in my earlier effort, but your example showed
me how much I missed the mark. That's why yo' da man ...

No, it's just basic by-the-numbers, but it does bring up a few points to remember.

1) This image shows the dangers of overreliance on monitor appearance. Certainly several of those who posted must have well-calibrated screens. And yet some of them posted backgrounds that are, upon examination, grossly blue or purple where the viewer expects approximate neutrality. This happens for the same reason that the camera's perception of the scene didn't agree with ours: we humans are very adept at calibrating our vision to the existing light. Look at one of these blue images for long enough, no matter how good the monitor, and it will start to look neutrally correct. That's why we need an Info palette.

2) Those people who still use master RGB curves or, worse, levels, might take this occasion to discard these primitive tools (or limit them to very small moves, where the harm they do is unlikely to be noticeable). This image is a spectacular example of why channel-by-channel is better, but even images that superficially look good can benefit from proper color handling, which master adjustments don't offer.

3) We don't know where this image came from, but somewhere along the line, either by human or by algorithm, it was damaged before we got it. By context, this is a dark image--at least dusk, maybe a true night scene. Yet the sky is extremely blue, even in the version I did. In many of the others, the sky is unprintably blue. This is one reason I did a conservative image--in real life we would have to go to extra effort to make up for the acquisition damage.

This effect occurs when somebody or some machine attempts to set a light point and blows out the lightest channels. This is a particular weakness of Camera Raw's automated adjustments, but it also occurs in some camera algorithms that lead to saving a JPEG. It's the type of image that may need to be *opened* in a raw acquisition module to avoid the damage that JPEGging may do (although, in fairness, many camera models wouldn't do it unless the identified the sky as *white*). OTOH, any attempt to do anything in Camera Raw other than a minimalist open would damage this image, because its corrections are applied, in effect, as a master curve.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Thu May 3, 2007 2:27 pm (PST)

The very first problem with all of this is the first statement "No, it's just basic by-the-numbers." None of us have any way of knowing what the *right* numbers should be unless we knew more about the intent of the photograph, but we don't. "Basic by-the-numbers" is a sensible concept in the context of a finished photographic file in the hands of a pre-press manager whose task it is to reproduce that image in offset as faithfully to the file numbers as the technology allows. More generally, "by the numbers" makes sense when the PURPOSE of the exercise is exact reproduction of the hues given by the Lab numbers. The concept does not necessarily make any sense in the context of photographic imagination, which is the present context. There is no presumption - apart from taste and intent or artistry - that the concrete should be any particular hue - it can be anything which is both believable and "atmospheric" for that time of the day. The hue of the sky in the original as presented was a not bad representation of what a sky could look like in the mind's eye at that time of the evening. And even if it were exaggerated a bit, if it looks good and believable it is good and believable - that is artistic license. More generally, unless you are dealing with a situation where the objects being photographed have known and precise colors and the client wants them reproduced exactly that way, image editing by the numbers is simply an extreme and not necessarily fail-safe approach to colour-balancing a photograph. Even when I photograph a Whi-bal and use it for adjusting images taken about the same time of day, I find myself departing from the cold logic of the numbers because the scene simply looks better, or more like I saw it, using some other recipe for the White Balance. *Accuracy* is a dangerous word in these situations. Very often, white balancing is a "mix to taste" affair. That may jangle some peoples' nerves because it leads to uncertain judgments rather than pat recipes, but that's life, and that's this image.

Now, if one wishes to handle colour balancing with artistic judgement rather than arithmetic alone, your first sentence in para (1) collapses, because in these conditions you absolutely need a properly calibrated and profiled monitor (which requires using a quality colorimeter and software), otherwise the adjustments you made to the numbers to produce what you like on the monitor will not come out that way from the printer, and you will be frustrated. Those of us using modern colour management techniques avoid such frustration - well, by using modern colour management techniques, the heart of which is a properly calibrated and profiled monitor on which we can place a high degree of reliance (complemented by the appropriate ICC profile governing the processing of the numbers from the profile connection space to the printer). You may still need an info palette, because in making these judgments it is often useful to have certain reference points for anchoring judgments. Those reference points would be colours whose Lab values you "trust", or to assess the extent to which a curve movement of "X" points will reduce a cast in the "a" or "b" channels. By working back and forth between the info palette and one's eyesight, believable and artistically meaningful images are produced. It isn't a case of either the numbers or the eyes, but use of both.

In making these adjustments, just to add a rejoinder on your comments about workflow, the place to start is a raw file opened in Camera Raw. As an experienced practitioner of raw file conversion using ACR, I would never employ automatic anything. All that stuff is turned OFF and I do it by hand image by image. Camera Raw in fact - especially in version 4 now shipping with CS3 (but even before) has superb tools for recovering image detail that may at first glance appear to be clipped. There is no weakness in Camera Raw per se - the only weakness may be on the part of inexperienced users who think they can depend on automatic settings to deliver them acceptable images. For technical reasons, those who know the inner workings of this module have advised that less damage is done to image data making the largest corrections in Camera Raw, then fine-tuning the image once opened in Photoshop.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: Andrew Rodney
Thu May 3, 2007 2:27 pm (PST)

On 5/3/07 7:10 AM, "dmargulisnj" wrote:

1) This image shows the dangers of overreliance on monitor appearance.
Certainly several of those who posted must have well-calibrated screens. And
yet some of them posted backgrounds that are, upon examination, grossly
blue or purple where the viewer expects approximate neutrality.

Says who?

This happens
for the same reason that the camera's perception of the scene didn't agree
with ours: we humans are very adept at calibrating our vision to the existing
light. Look at one of these blue images for long enough, no matter how good
the monitor, and it will start to look neutrally correct. That's why we need
an
Info palette.

So there's really no need to ever color correct an image, lets just adapt to the scene and move on.

3) We don't know where this image came from, but somewhere along the line,
either by human or by algorithm, it was damaged before we got it.

How do you know this?

By context,
this is a dark image--at least dusk, maybe a true night scene. Yet the sky is
extremely blue, even in the version I did. In many of the others, the sky is
unprintably blue.

Unprintable on what device? Every device on the planet? You©ˆve check this how?

This effect occurs when somebody or some machine attempts to set a light
point and blows out the lightest channels. This is a particular weakness of
Camera Raw's automated adjustments,

What automated adjustments based on what initial settings? You can roll your own. A particular weakness? Are you using this term to protect yourself from what is essentially an incorrect grossly exaggerated statement about ACR?

but it also occurs in some camera
algorithms that lead to saving a JPEG. It's the type of image that may need to
be *opened* in a raw acquisition module to avoid the damage that JPEGging
may do

May? Oh it most certainly will.

OTOH, any attempt to do anything in Camera
Raw other than a minimalist open would damage this image, because its
corrections are applied, in effect, as a master curve.

In effect? Is this your way of protecting the statement which is utterly incorrect? You keep saying this kind of unsupported nonsense but have yet to back it up making me suspect you don't know how to properly use this tool or how it applies its data corrections. Shall we once again investigate this using the products as designed? This idea of a master curve is absolutely not how the product handles the raw data, not even close. You really need to have a conversation with Thomas Knoll, Mark Hamburg (if they©ˆd even talk to you) or read Bruce Fraser©ˆs Real World Camera Raw. You are either ill informed about the processing pipeline or you©ˆre making this stuff up! You©ˆre entitled to believe anything you wish about this product but please don©ˆt give readers on this list the impression you©ˆre in any way correct about how this product processes raw data cause you©ˆre not!

Andrew Rodney
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Murray DeJager"
Thu May 3, 2007 2:27 pm (PST)

Hi Dan,

I've never participated in an exercise like this one before and I found it quite educational. It would have been even more educational if we had access to the "original" file, whether that be a RAW file or a Tiff file from a film scan. And maybe an "Accepted" final version as well.

Dan, you must have a hard drive full of such, real world, files. May I suggest that every now and then you toss one out for the masses to play with. Just like the one we just worked on, everyone who desires can pit there skills against one another and hopefully learn something in the meantime.

Murray DeJager
___________________________________________________________________________

Re: Color theory and practice
Posted by: Richard Wagner
Thu May 3, 2007 2:29 pm (PST)

On May 3, 2007, at 6:10 AM, dmargulisnj wrote:

3) We don't know where this image came from

True.

but somewhere along the line,
either by human or by algorithm, it was damaged before we got it.

Damaged? What, broken pixels? This is a meaningless statement, unless you are referring to the damage done by JPG compression. This image is not significantly clipped in either the highlights or the shadows, and the tiny amount of clipping that is present in the posted JPG is likely because of JPG compression, not from error by Boris in either exposure or processing. Please explain *exactly* what you mean by *damaged* so that we can follow your line of reasoning.

By context,
this is a dark image--at least dusk, maybe a true night scene. Yet the sky is
extremely blue, even in the version I did. In many of the others, the sky is
unprintably blue.

Unprintably blue on what device? We have no idea what the intended use of this image is. I can print the blues on Chromata Brilliance Canvas and a 9600 Epson just fine. Or on Hahnemühle FineArt Pearl. Or half a dozen other papers. You could make a 5-foot tall display print of this image with a decent inkjet printer and have no problems with the blues - with any of the versions presented. There is no reason to arbitrarily limit the gamut to that of, say, SWOP if the intended use is not known.

This is one reason I did a conservative image--in real life we
would have to go to extra effort to make up for the acquisition
damage.

What acquisition damage? Where is the evidence of this? What *specifically* are you talking about?

This effect occurs when somebody or some machine attempts to set a light
point and blows out the lightest channels.

Again, where is the evidence of this in the image presented? The ONLY thing that is "blown out" are the edges around the building - and this is likely an artifact of compression.

This is a particular weakness of
Camera Raw's automated adjustments

What?!!! There is no basis in fact to this statement. Furthermore, it is likely that Boris, obviously an experienced photographer, did not use the "automated adjustments" in Camera RAW (if that is indeed the RAW processor that he used, if he shot RAW) as anyone with any experience at all with RAW processing quickly turns off automated adjustments, regardless of which RAW processor they use. But regardless, Camera RAW does NOT "blow out the lightest channel" when used with the automated defaults. The automated settings are also not the "best" settings for making any given conversion, if that is what you do with your "open in RAW and save" technique.

, but it also occurs in some camera
algorithms that lead to saving a JPEG.

Camera processing to JPG is highly model/setting specific.

It's the type of image that may need to
be *opened* in a raw acquisition module to avoid the damage that JPEGging
may do (although, in fairness, many camera models wouldn't do it unless the
identified the sky as *white*).

There is always far more control in opening an image in RAW than in taking whatever processing the camera has done and saved as JPG, but for clarity, what JPG "damage" are you talking about? The reduction in colors in the original in adjacent pixels, or something else? The user has little or no control over the tone curve *or anything else* when the camera does all the processing and outputs a JPG.

OTOH, any attempt to do anything in Camera
Raw other than a minimalist open would damage this image, because its
corrections are applied, in effect, as a master curve.

This is pure, unsubstantiated, meaningless horse manure, and it displays a gross lack of understanding of the RAW processing pipeline, in ACR or any other RAW processor. First of all, what "damage" to the image by ACR are you talking about? This is a completely meaningless statement. Second, simply "opening" an image in Camera RAW does not give you an image to further manipulate in Photoshop. You have to SAVE the image, and when you've done that, guess what? You've done a conversion from a RAW image, with linear sensor data, to a gamma-corrected, output-referred color space (either sRGB, ColorMatch, AdobeRGB, or ProPhoto - your choice - but you have to choose one of them). You should get everything right BEFORE saving the file, whether that is exposure, white balance, tone curve adjustments, or anything else - it makes no sense whatsoever not to make these adjustments in Camera RAW. You can even optimize the conversion for whatever output space you choose, if you want to go that far.

To imply that a "better" image can be obtained by starting with a JPG, rather than a RAW image, simply does not jive with reality. The camera does the sensor-to-output conversion to make a JPG, but the user has essentially no control over the process. Even the simple adjustment of color temperature cannot be done as accurately on a JPG after-the-fact. With a RAW file, the demosaiced RGB channels have multipliers added to them *before* the image has any other processing done to it - and before it is converted from a scene-referred to an output-referred color space. Camera RAW can operate on the linear- encoded data, which you can't do with a JPG, that by definition is already in an output-referred color space. The list of examples goes on and on...

Contrary to your assertion that ACR is likely to "blow out" the highlights, ACR can frequently *recover* highlight detail by looking at a channel that has been overexposed (that's not ACR's fault!) and comparing other channel information in the vicinity to estimate what the correct values should be of the clipped channel. This technique can recover a surprising amount of detail, as anyone who has used it can attest. This is done on the linearly encoded data, not after conversion to the output-referred color space.

Essentially any correction to an image possible is best made on the RAW image, and not on the output-referred, gamma-adjusted pixel image. If you have evidence to the contrary, I (and a lot of other people) would like to see it. To state that *any* RAW processor is equivalent to using "master curves" in Photoshop is sheer nonsense.

--Rich Wagner
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Andrew S. Webb"
Thu May 3, 2007 6:54 pm (PST)

On May 3, 2007, at 7:10 AM, dmargulisnj arranged some pixels so they
looked like this:

OTOH, any attempt to do anything in Camera
Raw other than a minimalist open would damage this image, because its
corrections are applied, in effect, as a master curve.

What are you talking about here? Are you saying that Camera RAW corrections are applied in a manner siimilar to a master curve?

If that's what you're saying, you're wrong. Investigate CR further and you'll see why.

I'm sure you'll be relieved to hear that I agree with the rest of your post.

Cheers,

_andrew webb

P.S. See, Rodney and Wagner? You can disagree without being nasty.
___________________________________________________________________________

Re[2]: [colortheory] Re: Color theory and practice
Posted by: "Iliah Borg"
Thu May 3, 2007 6:57 pm (PST)

Dear Richard,

Thursday, May 3, 2007, 3:45:17 PM, you wrote:

Essentially any correction to an image possible is best made on the
RAW image, and not on the output-referred, gamma-adjusted pixel
image.

Now you need to list what corrections in ACR are applied to raw data.

--
Best regards,
Iliah Borg
___________________________________________________________________________

Re: Color theory and practice
Posted by: Stephen Marsh
Fri May 4, 2007 4:28 am (PST)

--- "Murray DeJager" wrote

I've never participated in an exercise like this one before and I
found it quite educational. It would have been even more educational
if we had access to the "original" file, whether that be a RAW file or
a Tiff file from a film scan. And maybe an "Accepted" final version as
well.

I asked Boris to look at a JPG of the original (offlist), as it may have been easier to enhance (globally or dropping in areas to his worked file) than the corrected image that was presented. The presented image to the group was quite a departure and the original was not much use as it was so far from the *intent* of the photographer's rendering. The presented rendering could be and probably is considered the original and the accepted final version - which we were asked to comment on.

Regards,

Stephen Marsh.
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Werner Tschan"
Fri May 4, 2007 6:09 am (PST)

Dear Andrew Rodney, dear Richard Wagner

You may be right, Dan may be wrong.
That doesn't justify the aggressive language you are using.
What's your problem?

Werner Tschan

--
STUDIO LTD
Atelier für Fotografie
Altenbergstrasse 8
CH-3013 Bern

T: ++41(0)31 332 88 33
F: ++41 (0)31 331 62 42
M: ++41(0)79 227 02 19
U: www.studio-ltd.com
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Lee Clawson"
Fri May 4, 2007 8:59 am (PST)

on 5/4/07 4:38 AM, Stephen Marsh wrote:

I asked Boris to look at a JPG of the original (offlist), as it may
have been easier to enhance (globally or dropping in areas to his
worked file) than the corrected image that was presented. The
presented image to the group was quite a departure and the original
was not much use as it was so far from the *intent* of the
photographer's rendering. The presented rendering could be and
probably is considered the original and the accepted final version -
which we were asked to comment on.

Stephen,

This sounds like I'm asking you when I would rather be asking this of Boris.

What I want to know if what's written above displaces Dan's speculation about acquisition damage, master curves and monitor appearance and puts it in Boris' hand ?

In other words can we assume that an image with edits that moved it so far from an original appearance mislead us ?? (To be fair Boris asked for an opinion. We later did image forsenics.)

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: Color theory and practice
Posted by: André Dumas
Fri May 4, 2007 9:01 am (PST)

Hello Dan,

Why do you say that it is a damaged image? Not because it is dark and the sky is blue (?) You mention the damage that JPEGging may do but I don't see any. Can you explain?

It is not correct to assume that concrete is gray (or neutral), that snow is white etc. under all lighting conditions yet your corrections seem to contradict that.

Boris' picture demonstrates that in the evening, and not only in daytime, the architect's design presents a well lit, pleasant and inviting façade. Those are important aspects in the design of a building and the picture shows that very well.

He would like an image that jumps out of the paper and I don't think that this is possible if it must also respect the overall mood that he chose for the image.

For my own satisfaction I assigned ColorMatch, did a Shadow/Highlight and felt that the image was improved. I also tried some curves but doing that merely moved the image away from Boris' original concept.

André Dumas
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Fri May 4, 2007 9:06 am (PST)

Andrew Webb writes,

What are you talking about here? Are you saying that Camera RAW
corrections are applied in a manner siimilar to a master curve?

Yes.

If that's what you're saying, you're wrong. Investigate CR further
and you'll see why.

No, I'm right. I tested the module carefully in preparation for PP5E. This feature particularly, because I was dubious that the implementation would have been done in the way it was.

Master-type corrections are undesirable because they increase and/or decrease saturation in ways that the user can't control. When a picture is being lightened by a master correction, saturation has to increase because the lightest channels are being hurt proportionally more than the dominating darker ones. Often that makes the image look better, but it's clearly the wrong policy to implement as a general rule--a serious user needs to make the decision of whether to saturate or not on a case-by-case basis.

In Boris's image, had it been in raw, I would *not* have wanted the blue skies to get bluer as the image got lighter. Camera Raw does not currently offer this capability in a graceful way, which is why I recommend that for best quality one should acquire in a minimalist fashion and then correct in Photoshop.

To illustrate, I grabbed the first raw image I found that had a dull, dark, blue sky of the type that would have been found in Boris's image if he had shot it raw. We would probably have wished to lighten such an image, but not make the sky significantly bluer.

A random point in the sky with Camera Raw's Exposure Control set to O and Adobe RGB specified for output measures 87r100g130b, or 42L(1)a(21)b. As it lightens, we do NOT want the B value to get more negative--we want it *less* negative if anything, with proportionally less lightening of the weak green and blue channels than of the darker red.

Instead, we get the master curve-like behavior of increasing the values in the three channels by roughly equal amounts, not percentages. When the Exposure setting goes to 1.5, the new values are 154r172g218b, which equates to the considerably bluer 70L(1)a(28)b.

By contrast, opening the original at 0 Exposure and then lightening in Photoshop with a luminosity, not a master curve, yields a much happier 154r167g197b.

If we had lightened using the Curve in Camera Raw rather than the Exposure setting, we would have gotten substantially the same result, The unwanted increase in saturation is characteristic of the master curve approach. Many cameras use this method internally, too, because it makes for brighter, happier colors. It is, however, a problematic implementation for a module that professionals are asked to use.

I'm sure you'll be relieved to hear that I agree with the rest of your post.

A load off my mind, indeed.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Fri May 4, 2007 1:59 pm (PST)

Mark Segal writes,

The very first problem with all of this is the first statement "No, it's just basic
by-the-numbers." None of us have any way of knowing what the *right*
numbers should be unless we knew more about the intent of the photograph,
but we don't. "Basic by-the-numbers" is a sensible concept in the context of a
finished photographic file in the hands of a pre-press manager whose task it
is to reproduce that image in offset as faithfully to the file numbers as the
technology allows. More generally, "by the numbers" makes sense when the
PURPOSE of the exercise is exact reproduction of the hues given by the Lab
numbers.

This misapprehends the purpose of color by the numbers, which is to establish a full range and eliminate impossible colors. By the numbers does *not* mean that we know what numbers are uniquely correct; it means that we identify certain combinations as defective.

Here, everyone agrees that liberties are to be taken with the original. Thousands of by-the-numbers-compliant versions are possible. For example, a variant of my image with a considerably darker, moodier three-quartertone would also be by-the-numbers correct. Some would presumably like it better and others would not.

Where there's a by-the-numbers error, *nobody* likes the bad version better. That's the case here with the greenish cast in the foreground concrete and the purplish one in the background. Give a jury a choice between one image with the two casts reduced and the other not, and the vote will be 100% in favor of the corrected version.

The hue of the sky in the original as
presented was a not bad representation of what a sky could look like in the
mind's eye at that time of the evening. And even if it were exaggerated a bit, if
it looks good and believable it is good and believable - that is artistic license.

Right. Skies, along with greenery and faces, are one of the three major areas in which we often desire extreme departures from the colors the camera saw. Commonly we prefer deeper and bluer skies in landscape shooting; in this image where the building is the product, I personally prefer a duller sky in order not to distract, but I have no problem with those who differ--provided that they stay within the range of a slightly blue cyan to a slightly purplish blue. People who do not check the numbers often hand in skies that are on the slightly green side of cyan. Such images are simply noncompetitive--the jury won't give them a single vote. *That's* by-the-numbers--not the specific sky you choose within the permissible range.

*Accuracy* is a dangerous word in these situations. Very often, white
balancing is a "mix to taste" affair. That may jangle some peoples' nerves
because it leads to uncertain judgments rather than pat recipes, but that's life,
and that's this image.

I have done a word search of my post and find no instance of either the word "accurate" or "accuracy". On the contrary, I have stated that there are many possible interpretations.

Now, if one wishes to handle colour balancing with artistic judgement
rather than arithmetic alone, your first sentence in para (1) collapses, because
in these conditions you absolutely need a properly calibrated and profiled
monitor

It is dangerous to use phrases like "you absolutely need" when the meaning is "*I* absolutely need". Some of the best color technicians around are in fact color-blind. In my advanced classes (not basic ACT) I force the students to make their creative judgments on a grayscale monitor. This year, as an experiment, I asked a subsequent class to identify, without a hint, which group of images had been corrected on a grayscale monitor and which on a color one, and the second class, which itself was quite expert, could not do so.

Such circumstances are admittedly artificial, but in today's world the better retouchers often have more work than time available to do it. It's not desirable per se to do color-critical work on a laptop in the middle of a commuter train with the light changing constantly and the laptop constantly changing position as the train bounces. But, we have to play the hand we're dealt. Those of us who travel a lot *have* to do work in somewhat uncalibrated settings whether we like it or not. We check the results on a calibrated monitor when possible, but if it isn't possible, it's no big deal. Only a laptop was available for the image that I posted to the group.

If the correct meaning of "you absolutely need" was "I absolutely need" then my advice would be to attempt to change that state of affairs.

In making these adjustments, just to add a rejoinder on your comments
about workflow, the place to start is a raw file opened in Camera Raw. As an
experienced practitioner of raw file conversion using ACR, I would never
employ automatic anything. All that stuff is turned OFF and I do it by hand
image by image.

Good.

Camera Raw in fact - especially in version 4 now shipping with CS3 (but
even before) has superb tools for recovering image detail that may at first
glance appear to be clipped.

That's correct. If we had a raw file to work with we could have protected the weaker channels and prevented the undesired saturating of the sky, if in fact we did not desire it. From what Boris describes, that saturating took place while the camera was producing a JPEG. With a raw file, we could reverse that unfortunate decision--but we'd have to accept a less than optimal luminosity, and correct it in Photoshop.

There is no weakness in Camera Raw per se - the only weakness may be
on the part of inexperienced users who think they can depend on automatic
settings to deliver them acceptable images.

The weakness is in any image that is not neutrally correct throughout, as this one is not. Here, I want to correct two similar near-neutral areas that are distinguishable only by darkness but have dissimilar casts that should each be minimized, without desaturating other areas of the same hue; plus I want to lighten the image overall without adding saturation. Camera Raw has no weapon that can do either.

For technical reasons, those who know the inner workings of this module
have advised that less damage is done to image data making the largest
corrections in Camera Raw, then fine-tuning the image once opened in
Photoshop.

I know they've said that, but they aren't correct.

FWIW, an interesting thread on by-the-numbers vs. calibrated monitor is going on at

http://www.dgrin.com/showthread.php?t=60276

Dan Margulis
___________________________________________________________________________

Color theory and practice - lessons learned
Posted by: "Boris Feldblyum"
Fri May 4, 2007 6:09 am (PST)

Folks,

This has been an overwhelming experience (39 age long file of all relevant emails) and I am truly humbled by it. No words can express my gratitude.

Reading this compilation just now, I felt like an imposter in a crowd of PhD’s. There is so much to learn! You have raised a wide range of issues and the least I can do is to answer as many of them as possible, hopefully on less than 39 pages.

As I mentioned before, I am an architectural photographer. All artistic and other aspirations aside, the main purpose of the photos is to help my clients (architects, interior designers, property managers, etc) to sell THEIR work to THEIR prospective clients. I am not dying to create dusk images because they are often striking , but only when it helps understand architecture. Certainly, a good image must be properly composed, with precise contrast and “believable colors”. This is my “comprehensive statement”. In this specific instance, my intention was to draw a viewer toward the theater entrance in the center of the image.

The (12) photos were delivered to the client yesterday. He did like most of them, including the one you dissected (“interesting colors, there is drama, I like it”).

Back to color theory. I have never studied PS systematically and my methods are home grown and are often crude. I do try to spend an hour every morning with Dan’s latest book if I am not shooting or on a deadline... My hat goes off to Davide D'Angelo who wrote: “I'm Italian and I've learned English reading this book”. I have been learning English for the last 30 years and I am still on Chapter four after as many months...

The photo was taken with Fuji S3, 12.5 MP. This is the only camera I now use, with the Nikkor 18-20 VR lens. I do miss the versatility of controls of a view camera and the depth of field of its lenses (the diffraction on the 18-20 lens creeps in at f11, and f16 or smaller is a no-no), but perspective can be controlled in post-processing and f8 is sufficient most of the time. (A new, “designed for digital” view camera plus a digital back is about US$30-40K). I use RAW when taking one-frame images, and the only control I change is the WhiteBalance. The files are then converted into 8-bit TIFs, using Hyper-Utility from Fuji.

The photo in question is a 3x4 mosaic, which produced a 22x42" or approximately 56x100cm @ 300dpi image. Several people, e.g. Steven Marsh, have suggested providing them with the original file. While this approach is the most appropriate one for image correction, I felt a bit embarrassed by outpouring of support and did not want impose more,

Please see illustrations on [
http: //www.bfcollection.net/errordocs/signature/signature.html
]http: //www.bfcollection.net/errordocs/signature/signature.html

Stitching is done by PanoTools, with PTMender which blends the exposures of all component images. If you are not familiar with the program, its worth checking out. Max Lions who wrote a GUI for the original software is an amazing guy and has a great sense of color. He also wrote a number of other useful image processing applications, e.g. Image Stacker. His web site is [ http://www.tawbaware.com/ ]http: //www.tawbaware.com/

Side note. Stitching is a very resource consuming process. On my old machine with single processor and 1GB RAM it could take 5-8 hours to produce a large a 3x4 mosaic. The new one (thank you group for helping me to evaluate the criteria), with Core2 Quad CPU @ 2.66GHz and 4GB RAM does the same in about 5-6 minutes. Additionally, it seems that PTMender is a faster than Enblend, a similar application which I used before.

The twelve component images were shot as JPGs for one reason: the camera records them faster than RAW, and it takes a few minutes to shoot a mosaic, during which time the light does change. I tend to underexpose in order to preserve highlights, but it was probably overdone here. Multiple exposures is another option but slows picture taking and blending is time consuming during post-processing.

(Side note. Dark horizontal lines visible on glass are the result of using polarizer. Tempered glass is a royal pain to photograph. It’s like writing curves. Reflections are reduced, lines appear; rotate the filter, reduce the lines, - reflections come back. Something to remember when taking photos)

The “best and final” image was cropped on the bottom as several of you have suggested. It is about 22x36in at 300dpi.

About color space. Please do not beat me hard but I use only sRGB. Yes, Adobe RGB is more appropriate, but there are other things to learn first. There is a pretty close match between my CRT and the printer (Epson 2200) so the set-up will remain for now. The clients like the look of Epson glossy prints. They do get the print quality files on CD/DVDs and make their own prints on wide variety of ink-jet and laser printers. Nothing matches the “original” Epson prints, but nobody complains. The published photos are very good in at least 90% cases, as far as I can can judge.

One other admittedly inefficient method I use (until I learn better) is to selectively correct different parts of an image. However I never forget that the entire image must be an accurate representation of reality so my brownstones will always be brown, not just the white horses. Unless they are supposed to be grey. (Are you confused yet? It’s intentional).

To sharpen images (always the last step) I create a duplicate layer, apply “sharpen more” filter and then go over the image with an eraser of various opacity, or simply reduce the duplicate layer’s opacity to my liking.

I cannot say now what exactly method was used to arrive at the “best and final” image. I tried to emulate what everybody has suggested and learned a lot. I used this image for training purposes and remember creating curves in all modes. Still do not know how to blend channels and what is a “blend if” method...

Please be kind if I missed your specific question. I will be happy to elaborate later.

Eternally grateful,

Boris
___________________________________________________________________________

Re: Color theory and practice - lessons learned
Posted by: "Murray DeJager"
Fri May 4, 2007 2:00 pm (PST)

Hi Boris,

Even before I looked at the color of your photo I was confused by what kind of camera and lens combination you had used to photograph this building. Now, reading below that this image was a mosaic makes sense.

The first thing I actually did with your photo was to use some "guides" in photoshop to try and figure out why some of the horizontal and vertical lines of the building didn't match. I wondered how one verticle line of the building could be perfectly plumb and not another. And at the same time different horizontal lines didn't seem to match in level. Or at least not in consistant manners like what would happen with wideangle lens distortions.

I raise this point because I personally found this very distracting while looking at the photo. I'd be curious if anyone else has ever raised this issue with you about doing this kind of mosaic work.

Murray DeJager
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Sat May 5, 2007 9:52 am (PST)

Murray writes,

I've never participated in an exercise like this one before and I found
it quite educational...
Dan, you must have a hard drive full of such, real world, files. May I
suggest that every now and then you toss one out for the masses to play
with. Just like the one we just worked on, everyone who desires can pit
there skills against one another and hopefully learn something in the
meantime.

Most learning about this field happens when we realize that our own versions are inadequate in some way, which as I recall is what your experience was with this particular image. Certainly it's the way that *I* improve my technique--if one of my students comes up with a version that's clearly better than the one I did, naturally I try to learn a lesson from it.

Doing that kind of exercise in the present list format is difficult because of the necessity to post the images offline. I note that John proposed a new exercise yesterday and got no takers. This is one reason for the recent suggestion of moving the list to a place where images may readily be included. The moderators have discussed the considerable response to that thread and are in the process of exploring whether the move can be made in a way that does not have the drawbacks listed by those who objected. Further news as it becomes available.

Dan Margulis
___________________________________________________________________________  

Re: Color theory and practice
Posted by: Dan Margulis
Sat May 5, 2007 1:15 pm (PST)

André Dumas writes,

Hello Dan,

Why do you say that it is a damaged image? Not because it is dark and
the sky is blue (?) You mention the damage that JPEGging may do but I
don't see any. Can you explain?

Actually, *you* explained. You tried to correct the image with curves and could not do so. This would not have occurred in the age of film--the picture is difficult to correct only because of how the data was manipulated before we got it. At the time I posted, we did not know where the damage occurred, but now Boris has indicated that the original capture was a JPEG.

Camera acquisition algorithms quite sensibly suppose that the typical user is not sophisticated and therefore do things that will give the best-looking results most of the time. Unfortunately, *some* of the time these same automated moves actually damage the image, giving us certain correction problems that did not exist prior to the age of digital photography.

Almost all images look better when they are given a full tonal range. So, algorithms always are reluctant to leave the capture flat--they stretch it out so that the lightest and darkest points are in fact very light and very dark. This move, if there was nothing more, would almost always improve the image.

Often, however, they go farther and deliberately discard a certain percentage of extreme values, or force the endpoints to be either white or black, or suppress highlight and/or shadow detail in the interest of more midtone contrast. These moves *usually* work, which is what an unsophisticated user wants, but on the occasions when they don't work, it's harder to fix the image afterward than if they hadn't been applied.

In lightening the light point the choices can be to discard roughly equal amounts from each channel, or roughly proportional ones, or somewhere in between. The first choice yields more intense colors and the second remains truer to the original flat-color feel.

If we are working ourselves on the flat image in Photoshop, of course we choose the method that gives the best result for the individual image. An algorithm, however, doesn't know anything about the image. And if we don't know anything about the image, then the first choice is probably the right one--we usually want brighter colors.

In a picture like Boris's, however, that move, in addition to making the sky bluer than it should be in context (which people might well accept) turns several small casts into large ones (which nobody would accept).

If you compare that original image to the unprocessed one, then it doubtless looks better--for the moment. But that's not the image that counts. The one that counts is the final one. If we start with a flat image there's no problem getting a good result. But with the additional saturation that the algorithm put in, Boris's original becomes an exercise for experts.

It is not correct to assume that concrete is gray (or neutral), that
snow is white etc. under all lighting conditions yet your corrections
seem to contradict that.

That human vision adjusts rapidly to the color of the ambient light (chromatic adaptation), resulting in the perception that neutral objects remain approximately neutral regardless of lighting, has been well understood for a bit over 200 years.

Chromatic adaptation (which a camera of course lacks) is but one of many factors which would fill several books but can be briefly summarized by saying that the human visual system is much more forgiving of variations in lighting than a camera is. A photographer may need a filter in a scene that a viewer does not find too bright, or a flash in a location that a viewer does not find too dark. So many books have been written on control of lighting for photography not because photographers have nothing better to do with their time, but because lighting variations that a layperson may think are trivial have a critical impact on the photograph.

The whole question of white balance in photography is engendered precisely because we humans *do* white balance the scene but cameras don't, not without some help. A great deal of research has been done on how to emulate the human white-balancing, resulting in at least two usable models.

Unfortunately, there is a lot of illogic in the human visual response that these formulas can't yet compensate for--the level of chromatic adaptation depends on the scene and on overall darkness among other things, plus we tolerate reproductions that are slightly too warm much better than we do those that are too cold.

Automated white-balancing is great when it works, but when it doesn't it engenders a category of image that is discussed at some length in PP5E: highlights and shadows correct, cast in the midrange. Such images were almost unknown in film, and they are much more difficult to correct. It was largely to avoid having to do so that I suggested putting something like Camera Raw into Photoshop before there was much awareness of digital photography.

Granted that, even with a good deal of scientific inquiry, we can't yet generate a method that automatically adjusts even for a single light source, it's no wonder that Boris's image, which has at least four, creates a problem. Not for us humans, of course: we just look at the scene and laugh it off. All we need is an algorithm that understands that the lighting is simultaneously all four things, each of which are simultaneously right and wrong, and each of which apply in different respects to different objects depending on context.

For my own satisfaction I assigned ColorMatch, did a Shadow/Highlight and felt that the image was improved. I also tried some curves but
doing that merely moved the image away from Boris' original concept.

These two moves, I am sure, would make the picture look better for almost any observer. If you in fact did not try to adjust the color further, I'd suggest downloading my image, placing it as a layer in Color mode on top of yours and, if you find the effect to disturbing, cutting opacity to 50%. Then ask yourself whether the image looks better with or without the top layer.

I did a couple of similar exercises with several groups some years ago. When the original casts were warm, the vote was split. When they were cold, as in Boris's picture, the vote was on the order of 700 to 0 in favor of moving toward neutrality.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Murray DeJager"
Sat May 5, 2007 8:27 pm (PST)

Hi Dan & John,

Quite the contrary! I played with John's image for over five hours last evening. After that I admited defeat and turned to a couple of books I have by Katrin Eismann for help. Nothing she had in her books on removing lens glare helped me much either. Actually, forgive me, there was one technique she had that involved converting the image to B&W and then basically painting all the colors back into the image manually that might have worked if I had a higher resolution image and more time.

But I did try. And I actually, once again, learned some things in the process; even though I failed at the original task.

Question: Dan, do you think this image is repairable?

Murray DeJager
___________________________________________________________________________

Re: Color theory and practice for ACR master curve
Posted by: "Mike Davis"
Sat May 5, 2007 9:53 am (PST)

I am way out of my league here, but it is my understanding that ACR and LR are essentially meta-data editors in which the original RAW (or now JPEG file) is left unchanged, while meta-data editing adjustments are placed between ACR/LR and the next editor or an output device. When the image is moved to, say Photoshop, the edited information becomes a part of the imported image for that session.

Photoshop, on the other hand, I see as a pixel-based editor which crunches pixels, either en-masse or by selection of areas, channels, or other chosen or manipulated grayscale values. Clearly, PS cannot handle RAW files directly, nor can ACR/LR deal with selections. ACR/LR allows us to use all the sensor data available and move the "window" to accomodate exposure latitude or error, but it does so on the meta-data, not on the pixel level.

It would seem that both are usually needed to maximize interpretive value of most images, and that individual control of channels is frequently essential to accurate color correction, aside from interpretive deviations.

--Mike Davis
___________________________________________________________________________  

By-the-numbers and "looks right" was (Case study: What to do about t
Posted by: Matthew Rigdon
Sat May 26, 2007 4:42 am (PST)

On May 25, 2007, at 6:08 AM, colorman042000 wrote:

Artistic interpretation aside, to the average onlooker, any image
will be by-the-number if it looks right. Period.

I'd change that to say that any image will be by-the-numbers if looks the way you intended it to (implying that you actually have to know what your intentions are to begin with).

If I decide, for artistic reasons, that my shadow point should be set at RGB (50,50,50) {Adobe RGB, so we don't fight over how to "interpret" the numbers} and that I want the entire image to have a red cast so I've personally decided that there should be two to three points more red in my "neutrals" because I find it best evokes my sense of "Kodachrome" in my viewers, so long as the image meets those standards it will be by-the-numbers compliant.

Of course, if you had no idea that I had made these artistic decisions, then you would look at my photograph and claim it's wrong. Which is why you either need to know what I am trying to do with the photograph or you need to have seen a substantial portion of my portfolio so you could have determined that I don't like dark shadows and all of my photos have a reddish tinge to them.

Because I can go from your definition of by-the-number wrong to by- the-number right just by communicating to you what I'm doing (or by you taking the time to familiarize yourself with my work, although I don't actually do this, but maybe it would work. As a photographer, you do want to have a visual style that will set you apart and if you have the exact same or very close to the same style as everyone who took Dan's class, you may have to look at other things to make your mark).

But I don't think that by-the-numbers has ever claimed to make things LOOK right. It just claims to make the numbers what you chose to make the numbers. And if you're any good at choosing your numbers, you images will look right. Whatever right may be.

Matthew Rigdon
http://techsonnet.blogspot.com
___________________________________________________________________________

Re: Case studies?
Posted by: "Henry"
Fri May 25, 2007 11:02 am (PST)

On May 23, 2007, at 4:58 PM, MARK SEGAL wrote:

Actually, working by the numbers can be a very bad thing if it
messes-up the impression the photograph was intended to convey. The
term "trustworthy" may apply - for example - to photographs of
sweaters for a Lands End catalogue, but in fine art photography it
just isn't relevant.

That was one of the directions I was aiming with my point about there being a distinction between the recorder of an image and a creator of an image. A fIne art photograph implies that there is one person who is both the recorder and creator of the image.

Some artists strive for uniqueness by avoiding attributes in their art that are considered "normal". In photography, a BTN approach might actually be helpful if one is attempting to purposefully avoid "normal-ish" numbers. In painting, one begins by learning normal or correct relationships before progressing to work that is less representational. I would think that the same would apply to photography.

A representational image can be vetted for trustworthiness by using a BTN approach.

Henry Davis
___________________________________________________________________________

A fictional case study
Posted by: Matthew Rigdon
Sat May 26, 2007 6:50 am (PST)

To take a page out of Professional Photoshop, I’ll offer up a fictional case study that I believe would stump the bulk of the list members.

Take a medium shot of a bald caucasian male suffering from jaundice in front of a brownstone wall.

“I flew to New York and took this picture of my dying friend. Anyone have any suggestions how to correct it? I want to have a picture to remember my last visit printed up 8x10.”

I would wager that 98% of the corrections submitted would be WRONG. I’ll allow that there might be a member of the list who lives in New York, who walks by this particular building every day, and recognizes something in the pattern of the bricks or a particular feature of the wall itself who would say, “This wall is brown. This is what I would expect. So I guess I have to believe this man is yellow. Well, he is dying. I guess he could be dying of liver failure.” I’ll call the odds of this at slim-to-none.

There might be a particularly intrepid member of the list who, having downloaded the RAW file, might interrogate the EXIF data from the camera and determine that the picture was taken on January 27, 2007 at 1 PM. He might then determine from the records of the weather that day that it was a clear winter day and, relying on his long experience with photography or by consulting the appropriate sources, determine that the camera’s white balance circuitry recorded the wrong white balance point and then plugin the appropriate number in ACR, then correct the photograph and still be WRONG.

You see, there was another even more intrepid member who went to the photographer’s webpage and clicked on the About Me link and read about how the photographer is a resident of Seattle and then noticed that the picture was taken on January 27, 2007 at 1 PM Pacific Daylight Time, so it was actually 4 PM in New York and cloud cover had drifted in during those intervening three hours and it turns out the camera’s white balance circuitry was functioning properly and that the wall is brown and the man is yellow.

Although there’s still the possibility that he (because he read Dan’s book, knows that some walls are brown, especially in New York, but he still cannot believe that the man is yellow because he’s never heard of jaundice) might choose to mask the man out and correct him to a healthy pinkish hue.

We could vote and debate and argue and come to a “conclusion” and all we will have determined is whose photo had the best contrast. All of the color is wrong, except for the two cases above, which would probably have been voted off the island in the early stages because “Walls are gray” and “Caucasian men aren’t that yellow.”

Which is to say, in the absence of the appropriate information, anything you do to a photograph is meaningless.

And while the exercise, if actually carried out, would be a colossal waste of time (other than to teach all of you to never trust a word I say), it is very instructive if you truly want to understand the system. It is only sometimes instructive to study a system when it works comfortably. It is ALWAYS instructive to study a system when and where it breaks down.

Or to borrow from Frank Herbert, I’ll just quote the Amtal Rule (from the Dune series of novels):

“To know a thing well, know its limits. Only when pushed beyond its tolerances will true nature be seen.”

If you choose the wrong numbers (and there is always choice), your corrections will always be wrong.

For those of you who would object that my request didn’t provide enough information, I can also give too much information.

“This is my friend Brad. Brad and I served in Vietnam together. One day, on a patrol boat going down the river to find Colonel Kurtz, Brad threw himself onto a grenade that was thrown into our boat. He had to be shipped home and spent 18 months at Walter Reed. It was a miracle he even survived.

“While he was there, he developed an addition to morphine. It took him another five years to kick that, but he picked up an addiction to alcohol. He’s been an alcoholic for the last thirty years. It cost him his career (his father ran a successful investment fund on Wall Street), both of his marriages, and now, his liver. The doctors tell us he’ll be dead before spring. He’s ineligible for the transplant because of his age and his alcoholism.

“Despite all this, I love him like a brother. I’d literally give my life for him, just as he did for me. I flew out to see him, probably for the last time ever as I’m about to go on an extended tour of Canada, so I wanted to get one last picture. I hope to blow this up to 8x10 to hang up in the living room. I’d appreciate any help or any suggestions for correcting this.

“Thanks.”

See, I just told you he has jaundice. How could you miss it?

Matthew Rigdon
http://techsonnet.blogspot.com
___________________________________________________________________________

Re: A fictional case study
Posted by: Matthew Rigdon
Sat May 26, 2007 9:13 am (PST)

I totally forgot: he's wearing a New England Patriots jersey from the 1985 season when they went to Super Bowl XX. I bought a brand new Chicago Bears jersey and used a paint pen to write the score of this past Super Bowl on it, and while it made him laugh, he still refused to wear it.

And it doesn't demonstrate that by-the-numbers is BS. It demonstrates that no one can correct this image, whether they work by-the-numbers, whether they convert to ProPhotoRGB, or they use an EyeOne Display to calibrate their Samsung LED monitor every single morning, whether they run it through Auto Color, Auto Eye, or tweak a luminance curve in Lightroom. No one can correct this photo if they don't know:

This man has jaundice.

It's like a riddle wrapped in a conundrum. Mixed with secret sauce.

Matthew Rigdon
http://techsonnet.blogspot.com
___________________________________________________________________________

Re: A fictional case study
Posted by: Dan Margulis
Sat May 26, 2007 2:51 pm (PST)

Matthew writes,

Take a medium shot of a bald caucasian male suffering from jaundice
in front of a brownstone wall.
"I flew to New York and took this picture of my dying friend. Anyone
have any suggestions how to correct it? I want to have a picture to
remember my last visit printed up 8x10."
I would wager that 98% of the corrections submitted would be WRONG.

Many people would indeed get it wrong, but an experienced BTN practitioner would not.

First, it sounds like you've had some recent experience with someone dying of this, as have I. I'm sorry. If you visualize that person, you'll recall that his skin was not a uniform yellow, but rather mottled. Furthermore, it is sadly obvious that the person is in failing health. So, the operator has to be alive to the possibility that normal fleshtone values may not be present.

There is a similar example in PP5E of a congressman teaching a group of schoolchildren. His face is extraordinarily pink--as proven by other verifiable numbers in the image.

I'll allow that there might be a member of the list who lives in New
York, who walks by this particular building every day, and recognizes
something in the pattern of the bricks or a particular feature of the
wall itself who would say, "This wall is brown. This is what I would
expect. So I guess I have to believe this man is yellow. Well, he is
dying. I guess he could be dying of liver failure." I'll call the
odds of this at slim-to-none.

I use a New York City brownstone in my ACT classes for this very reason--it traps people into deciding that they know for a fact that it's gray, even though there are ample other known elements in the image that prove it cannot be. That is the case here, too.

If I were not from the New York area, I would regard it as unlikely that the wall was anything other than a gray, or slightly blue off-gray, or slightly warm off-gray. This is not enough to *force* it to be one of these colors, as I told André when he said he was 95% sure as to the color of the ballerina's slippers.

If I do not know that the person has liver failure, then I would regard it as *very* unlikely that his skin is as yellow as it is. But consider the alternatives.

1) If the wall is forced to be gray or anything close, then the skintone must become green, as the red channel darkens. That's not "very unlikely". It's impossible. Therefore the wall is not neutral, which means it must be some form of brown.

2) Brown is a species of red. We expect roughly equal values in the green and blue channels, or in the magenta and yellow channels, with the red/cyan channels being lighter. In the event that the two darker channels are not equal, then the blue/yellow must be darker, creating a yellowish brown rather than a purplish one.

3) But if the skin is forced (by darkening green/magenta) to conform to a conventional fleshtone value, the wall becomes very purple. This is not "very unlikely." It's impossible. Therefore the fleshtone value cannot be normal. There is room for debate as to how yellow the fleshtone needs to be, but not that it should be yellow.

As an eminent color scientist once remarked, "When you have eliminated the impossible, whatever remains, however improbable, is the truth."

Dan Margulis
___________________________________________________________________________

Re: A fictional case study
Posted by: Matthew Rigdon
Sat May 26, 2007 5:45 pm (PST)

On May 26, 2007, at 11:46 AM, dmargulisnj wrote:

First, it sounds like you've had some recent experience with
someone dying of this, as have I. I'm sorry. If you visualize that
person, you'll recall that his skin was not a uniform yellow, but
rather mottled. Furthermore, it is sadly obvious that the person is
in failing health. So, the operator has to be alive to the
possibility that normal fleshtone values may not be present.

Fortunately, no. Some years back, I had a friend whose son had to stay in the pediatric wing at UCLA. One of the other patients was a little two-year-old girl with congenital liver failure. I did remember the sickly yellow color, but not as much the mottling of skin (mottling could happen for other reasons), but it does make your sclera yellow which would be another tip-off that the man in the picture was suffering from jaundice. The eyes were what I remember most (this particular girl was Hispanic, so her skin tones would have been different than the man described).

Since he's an older man (he served in Vietnam) and it's also a winter's day in New York, someone might attribute the mottling to burst blood vessels in the skin from wind chill or from the use of alcohol over the years.

1) If the wall is forced to be gray or anything close, then the
skintone must become green, as the red channel darkens. That's not
"very unlikely". It's impossible. Therefore the wall is not
neutral, which means it must be some form of brown.

A quick trip over to Google tells me that green skin is sometimes associated with multiple organ failure and if all you knew is that my friend is dying, you might assume his skin is green. Now, you could wonder how a man dying of multiple organ failure got outside to take a picture, but I should say that I use the definition of medium shot meaning it's a shot of him from the waste up. So he could be sitting down on a bench. Let's say I lifted him out of his wheelchair because I didn't want the wheelchair in the picture (I only want to leave the wheelchair out because I don't want to give you any reliable way to determine the color).


2) Brown is a species of red. We expect roughly equal values in the
green and blue channels, or in the magenta and yellow channels,
with the red/cyan channels being lighter. In the event that the two
darker channels are not equal, then the blue/yellow must be darker,
creating a yellowish brown rather than a purplish one.

As you demonstrate, if you know for certain that the wall is brownstone, you can correct this photo properly. Although you would still need to figure out if someone's skin could be yellow to keep you from doing something crazy to "fix" him.

3) But if the skin is forced (by darkening green/magenta) to
conform to a conventional fleshtone value, the wall becomes very
purple. This is not "very unlikely." It's impossible. Therefore the
fleshtone value cannot be normal. There is room for debate as to
how yellow the fleshtone needs to be, but not that it should be
yellow.

I won't say impossible, but it's only because I grew up in Fort Worth, Texas. You see, some years back the city of Fort Worth built a new courthouse. The architect chose to design a thoroughly modern building, inside and out. If you've been to Texas, you know that the old courthouses in that state are very distinctive. The people of Fort Worth hated this new building, but the city couldn't afford to tear down the new courthouse and rebuild in the style of the old one, so they paid a company to erect a gigantic faux facade that, if you don't get too close, makes the new courthouse LOOK like a traditional courthouse made out of pink granite and marble. But in reality, it's just a bunch of wood that's been painted pink. So what color is the wood now? Plus, they've figured out ways to dye concrete. So if I bet my image on the man skin's and I bet that the skin was a "normal" color, there could be a purple building somewhere in New York (I don't live in New York, but I do live in Los Angeles and we have purple people, so I have no trouble believing in a purple wall).

But let's say I even tell you that I carry an 18% gray card every where and shoot it before every series of pictures I take and go into the menu on my 20D and load a custom white point. So the jpeg I submitted shows a yellow man standing in front of a brown wall. How many people would believe what they see? I've even told you that I made sure to capture the exact white point for the scene, so the camera didn't get confused. Who would believe their eyes? And I don't believe there's even any software that could "correct" this photo, as I don't know that any correction software can perform medical diagnostic tests on the people in the picture or have any clue that he's standing in front of brownstone (and what if bought some of that fake stone fluffy paint from Michael's and sprayed it on a piece of plywood, then what would the software do?)

So if you pick your colors wrong, your correction will be wrong (but there are lots of logical reasons why you might pick wrong colors) and I wonder how many of you would be willing to bet the image on the camera's data. If I gave you no description of the scene, but just the photo, many of you who "trust" the electronics might assume that I screwed up the custom white balance I set.

Matthew Rigdon
http://techsonnet.blogspot.com
___________________________________________________________________________

Re: A fictional case study
Posted by: Gene Palmiter
Sun May 27, 2007 3:45 am (PST)

Firstly! I resent the implication that Vietnam Vets are 'older men'. I joined up at the age of 17 in December of 1971. I turned 17 the previous November. I am only 52...not an older man! Secondly, not all Vietnam vets have drinking problems.

In one (or more) of Dan's books there is a photo of a wine label that illustrates the problem of making false assumptions about what color things ought to be.
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Dan Margulis
Sat May 26, 2007 2:52 pm (PST)

André writes,

I wonder if Dan would agree to post his version (shown in Figure 3.2)
so that we can evaluate it against our own and the original (on the CD).

The curves that produced this image are shown in the text of the book. Being an introductory chapter, nothing else was done to the image--no sharpening, LAB, or manipulation in CMYK. There is no printing issue with respect to its reproduction.

Owners of CDs from either my books or my classes are reminded that the license agreement permits only private, personal use of the images for educational purposes only. Posting, sharing, or otherwise publishing them is prohibited. This is for the benefit of the photographers who so kindly permitted use of their work. They did not authorize me or anybody else to launch them into cyberspace.

Artistic interpretation aside, to the average onlooker, any image
will be by-the-number if it looks right. Period.

Unfortunately, even experts, let alone average onlookers, can't tell whether their images look right unless they see what others can do with them. John referred to this effect earlier when he proposed competitions. It's as true of me as of him: so many times I've prepared an image thinking it was good, only to have its defects exposed by three other people who did a better job. Look again at the Giselle images, which have little to do with BTN. Some of what was posted was clearly poor, and the people who posted it would undoubtedly be the first to admit it, now that they have something to compare it to. But surely it must have "looked right" to them when they *posted* it--and members of this list are considerably more sophisticated than the "average onlooker"

The inability to see what's wrong with an image is particularly pronounced with respect to BTN. We saw that here, with Boris's image, where at least two list members said things like "I can't believe I was as far off as that" when they saw what proper BTN analysis could do--yet the image must have "looked right" when they posted it.

But there's an even better example: my class attracts a fair number of very prestigious professional photographers, people who have a national or international reputation for quality. It is not unheard of for them to be good, but in the large majority of cases, in spite of their reputation, they are near the bottom of the class in terms of ability to improve an image. The problem is not lack of artistic vision, but the exact opposite: too much of a willingness to trust their own instincts rather than to check the image in a disciplined fashion. As a result, on the first two days of the class their results are consistently equalled, and often surpassed, by beginners.

People do not reach that level of recognition by being stupid, and these photographers don't take very long to decide that having beginners beat them in image quality is an unsatisfactory state of affairs. So by the end of the three days all is well except that, being as quality-conscious as they are, they often go back and rework a ton of images that they had previously released as "looking right". I have been shown images that won *awards*--and the photographer now says in retrospect, "I can't believe what a POS this is!" When that happens, it's almost invariably because he (and the judges) thought that the picture "looked right."

Dan Margulis
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Chris Murphy
Sun May 27, 2007 3:44 am (PST)

There are three functions for color correction:

1. Special effects, compositing that alter the image away from the original capture. This can also constitute a "look" that a particular photographer wants their images to have, even if the original scene or the capture don't reflect it (yet).

2. Getting a good rendering of an image. This accounts for the fact the capture has a wider gamut and dynamic range than print, and something has to be compromised in order to make it "look good" in print. If you know what it's supposed to look like, the easier the job is. On the color science end, this is referred to as color appearance modeling and there has been very sophisticated work, beyond that of just colorimetric matching (which is mostly what color management has been trying to do thus far.)

3. Getting the numbers right to preserve color appearance you want, but accounting for the fact different devices and media produce different colors for the same numbers. This is the predominant area where color management has ruled - in the context of mapping output- referred data to various output devices.

The vast majority of correction being talked about in these threads of late are about getting a good rendering. Getting an image that looks good. In many cases, people have been correcting images to unwind previous rendering attempts by either the film being used itself (if its "look" wasn't exactly what was desired), or today with digital cameras and associated software that bake in a rendering that the user can't control. This is found with JPEG and with the default settings with most raw converters. Lightroom is mostly an exception because its rendering is rather flat. That does't make it a pleasing default rendering but it's not intended to be one. Rather it's intended to be a starting point so you can see what's been captured, and what you have to work with. The job of rendering the image is being left up to the end user to determine.

JPEG from cameras aren't inherently inferior to RAW, they're just different. The thing to know about JPEG is that it has a lot of camera manufacturer rendering baked into it. They are shifting colors and tone all over the place in an attempt to get what they think is a good rendering for a particular image. The contortions they're putting images through, increasingly, are significant. There's clipping at both ends, and compression is even occurring in midtones. Colors are being stretched and pulled. Skin tones are being recognized and altered. There is no choice in the matter at all. You can overexpose and underexpose thinking maybe you won't clip shadows or highlights but these renderings are very aggressive. You don't know what's being thrown out because it's not documented.

RAW is unrendered. It represents scene-referred data that can have upwards of 8000:1 contrast ratio, so the dynamic range is quite a bit bigger than that of printing which will have at best around 250: 1 and in many real world situations far less than that, less than 100:1. The technique of cramming an 8000:1 capture into a 100:1 printed piece has been in large part considered an art form. RAW converters including ACR are a tool for user definable rendering. You decide what is clipped, compressed, saturated, desaturated, etc.

It's not always easy to distinguish between rendering and artistic modification. This is because the controls for producing a good rendering also allow for artistic modification. ACR obviously allows you to some radical things to an image. But a good rendering simply looks good and MOST people will agree on what is the best rendering when they see it compared to not merely other options but the original scene. And only those privy to that original scene are really well suited for saying what is a good rendering for that scene. They may not be able to get there themselves, but once arrived at, they'll recognize it at a rate that is by far not random.

Now, getting good renderings is a function of color appearance models and quite a bit more is known about what makes a good rendering than has yet been employed by any piece of publically availalbe software thus far. But I am hopeful that soon we will see the first generation of some of these products, employing color appearance models that take things like luminance and color temperature into account. More sophisticated color appearance models can take partial chromatic adaptation into account as well. So I predict we will see software come up with better default renderings out of the gate using the information already contained in the RAW file that is not even remotely contained in a JPEG. And that will be a much better starting point from which to further enhance or modify the image, rather than having to under some other rendering that's previously been applied.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Andre Dumas
Sun May 27, 2007 4:56 pm (PST)

Hello Dan,

Unfortunately, even experts, let alone average onlookers, can't
tell whether their images look right unless they see what others can
do with them. John referred to this effect earlier when he proposed
competitions.

When I decide that "this one is the one" then I can tell that my image does *look right*. Later, when I see what someone else can do with my image I may take it as just another interpretation or I may like it better than my own. This is not a special phenomenon ( "effect" ) and it does not mean that the original was not right.

The inability to see what's wrong with an image is particularly
pronounced with respect to BTN. We saw that here, with Boris's image,
where at least two list members said things like "I can't believe I
was as far off as that" when they saw what proper BTN analysis could
do--yet the image must have "looked right" when they posted it.

I don't think that you can define what is wrong or right with an image by analyzing it against a rule called "By The Numbers". You have defined BTN, but when I see how you apply your definition to different images then I don't agree with it.

André Dumas
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Chris Murphy
Sun May 27, 2007 5:03 pm (PST)

On May 26, 2007, at 2:51 PM, dmargulisnj wrote:

But there's an even better example: my class attracts a fair number
of very prestigious professional photographers, people who have a
national or international reputation for quality. It is not unheard
of for them to be good, but in the large majority of cases, in
spite of their reputation, they are near the bottom of the class in
terms of ability to improve an image. The problem is not lack of
artistic vision, but the exact opposite: too much of a willingness
to trust their own instincts rather than to check the image in a
disciplined fashion. As a result, on the first two days of the
class their results are consistently equalled, and often surpassed,
by beginners.

*Shrug* there's aways a propensity for those who have been doing something a particular way for a very long time to assume that their way of doing things is incontrovertible.

A prestigious professional photographer has likely been so for some time. And arrived at that position before they had the tools and control to get a good rendering. For years they've been used to getting prebaked renderings from film, and now with JPEG, if that's been their workflow.

With film only it was simply they got what they got. Then Photoshop and scanning came along and they could undo some aspects of the manufacturer's rendering, and get something else. And now with RAW they can define and apply their own rendering out of the gate. The skill sets for these things are different.

People do not reach that level of recognition by being stupid, and
these photographers don't take very long to decide that having
beginners beat them in image quality is an unsatisfactory state of
affairs. So by the end of the three days all is well except that,
being as quality-conscious as they are, they often go back and
rework a ton of images that they had previously released as
"looking right". I have been shown images that won *awards*--and
the photographer now says in retrospect, "I can't believe what a
POS this is!" When that happens, it's almost invariably because he
(and the judges) thought that the picture "looked right."

We should be very clear on differentiating good renderings from artistic deviation from reality. The radically altered piece isn't always going to get a majority of people saying "ooh yeah nice." It's not designed to. In either case, the original photographer is the final arbiter of what is a good rendering of the original scene, since he was there, as well as any artistic modification taking the image beyond that point. You can like it or not, but you can't rationally talk about art as being right or wrong.

So, let's not extrapolate beyond the singular "POS" event. What you're characterizing here is not someone who can't identify a good rendering when he sees it, because clearly he did recognize it. What seems obvious to me is that this photographer simply wasn't very good at manipulating images in Photoshop to arrive at a good rendering, a good rendering that he identified with immediately. Perhaps he should spend more time shooting and evaluating images, and hire someone who is adept at translating his vision into the correct Photoshop maneuvering.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Matthew Rigdon
Sun May 27, 2007 5:03 pm (PST)

On May 27, 2007, at 12:09 AM, Chris Murphy wrote:

But I am hopeful that soon we will see the first generation
of some of these products, employing color appearance models that
take things like luminance and color temperature into account. More
sophisticated color appearance models can take partial chromatic
adaptation into account as well.

The real question is, when will we really see this technology? But I'd argue that there are many things you can do, especially if you're the photographer, that have nothing to do with the technology that will improve your work. I don't know that any software will ever achieve what a disciplined human can do. But I'd argue that you have to teach yourself how to see. It’s something you have to train yourself to do.

I put forward that digital photographers in particular, but anyone in general, often fall prey to certain fallacies that sabotage any attempt they would make to get accurate color. The fallacy is this:

The camera captures a color image, therefore the camera causes color.

I’ll offer a little story.

So I go to church this morning at 6 A.M., which means I’m probably a little tired and loopy, but putting that aside, this particular church has a couple of screens hanging up on either side of the stage on which they project a video feed of the proceedings. You see, the church meets in what was once a warehouse which now holds about 2500 people, so if you sit in the back, it’s very hard to see who or what might be on the stage. I notice this morning that the colors of the robes of the choir members and the skintones of the various people that I see don’t look anything like what’s being projected on those screens. It’s happening in real time, so as the camera pans, I can go back and forth and measure, with my own eye, and see that the camera is shooting another choir in some other dimension. First, I peer a little harder at the projector screen and I figure the screen is old, it’s probably discolored, and if they would just buy new projector screens, the color would be correct.

Then I decide that these must be older LCD projectors, so they were probably using less accurate fluorescent bulbs to provide illumination, so if they’d only replace these projectors with newer models that use some sort of LED lighting, the color of the choir on stage and the color of the choir in that video would match.

But now I recall how video engineers have always joked that NTSC stands for “Never Twice the Same Color” and I only think that if we could convince the powers that be to replace the NTSC standard with something better, the color would be fixed.

But then I remember the cameras. You see, when I walked in I noticed a camera operator with a Canon XL-1s standing a few rows in front of me. That’s why I moved over three seats. I figured that the church was getting by with older equipment, such as the LCD projectors I blamed earlier, because I know that Canon doesn’t manufacture the XL-1s any more. There’s another Canon XL-1s on the other side of the auditorium and they both appear to be outfitted identically, so it appears that these two cameras should capture same image or some semblance of. And the video feed cuts between the two and the cameras seem to be recording the same choir, but it clearly isn’t the choir thats performing on stage. It’s a completely different color. I recall that Canon equipped these cameras with a set of three (3) quarter-inch CCD chips and that these cameras sample the green CCD (but only the green CCD) twice every field in order to create the final image, a technique that Canon’s engineers dubbed “pixel- doubling”. They claimed this provided the same level of performance as the larger one-third-inch chips that Sony and Panasonic used in their competing cameras. I know that this caused a discernible color cast to the footage that makes it problematic to cut footage from a Canon camera in with footage that came from a Sony or a Panasonic. All of the Canon footage would have to be color corrected first. So now I’ve decided that if the church would only replace the Canon cameras with something from Sony or Panasonic, the color would be correct.

Now, and this is the point during the service where I finally fall down the right rabbit hole, although I don’t realize it yet, as I look up over the stage and I notice scaffolding. There are over two dozen stage lights attached to this scaffold in various places, pointing to different locations such as the pulpit, the choir, the band, and the piano. I can tell from the design of the lights that they have some bulb that works on the principle as the familiar tungsten lamp, but because I used to work for a company that manufactured all manner of lightbulbs, I know that the filaments in these lamps are doped with all sorts of exotic substances in order to provide longer life or lower the energy lost to heat, and that all of these substances change the emissive properties of the lamps in ways that the engineers probably didn’t care about. As the service goes on, I note that the lights fade up and down according to the song or speaker and sometimes these lights appear white and other times they appear somewhat yellow. And I’m so fascinated by these lights that I notice that one of the lights pointing at the stage has a purple gel on it. Then I see there’s another light that has a pink gel. And over there, to one side, are three lights that have orange gels. And I wonder if any of these lights are pointing out at the audience, but they aren’t, though a couple are set up to function as backlights for the pulpit speaker, but this is when I notice that they’ve hung up halogen work lamps over the congregation in order to provide us with light to see. And as I look out over the audience, my eyes are drawn to the paintings that are hung up every ten feet on the walls. Each of these paintings is lit with what appears to be a SoLux lamp with barn doors.

And at this point I realize that it’s a miracle that what I see on the video screen looks anything like the service I’m physically attending. When I look up to thank whatever lighting gods may be, I utter another prayer thanking them that the fluorescent tubes overheard are shut off and there are no windows, as otherwise the lighting would have created a wedding-at-Cana style miracle again and turned the pastor’s wine colored suit into something other than gray on the video screen.

Which is to say, the hunk of electronics, magnesium alloy, and precision ground glass you’re holding in your hand has nothing to do with color.

I even know, by accident, that I can completely screw up the auto white balance of my Canon in reproducible ways. One night, when I was just playing around with some settings, familiarizing myself with the camera, I took a picture of my apartment. I even shot it in RAW format. When I opened up the picture on my computer, I was mystified. You see, I have a halogen torchiere lamp on one side of the room and a plain old table lamp on the other, but I’ve replaced all my old tungsten bulbs with these new fluorescent bulbs that save energy. And the room looks fine. I mean, the fluorescent light is a bit “off” at first, but I convince myself that I’m saving money and it’s okay. And it is. My room still looks the same. It still looks the same even if the halogen and fluorescent lights are both on.

But gosh-darn-it, every time I take a picture with this stupid camera, it shows me another world. So I figure one of two things are happening. One, the camera’s CMOS sensor is doped with some substance that causes my room to become a magical land of tangerine trees and marmalade skies. Two, when I activate the shutter, there is some sort of electrical field generated by the lithium-ion battery, the halogen filament, and the fluorescent gases that causes my camera to momentarily wink into some n-dimensional version of my apartment, snap a picture, and then return to my hand in this universe.

The funny thing is that the camera, Adobe Camera Raw, and Aperture are all absolutely, positively convinced that my room is lit by a fluoresogen bulb and framed by a Las Vegas revue. And for the life of me, I don’t know how to convince them otherwise.

My camera is magical.

Of course, many people with digital cameras often believe this because, well, they never stop looking through the viewfinder and pressing the shutter. They figure that if they take enough pictures, one of them will be correct. And that there couldn’t possibly be anything outside of a 12-micron radius of their camera body that could be affecting the color. And that if it cost enough money, it surely must always get the color right.

If you know TOO much about cameras, you’d might even tell me that if I had a Fuji with it’s dual-sized pixels, or even a Sigma SD-14 with the Foveon sensor, the problem would go away. It wouldn’t. Those cameras might transport me to yet another world, perhaps a world in which purple haze gets in my face, but it still wouldn’t look like my apartment.

Matthew Rigdon
http://techsonnet.blogspot.com
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Chris Murphy
Mon May 28, 2007 4:00 am (PST)

On May 27, 2007, at 1:21 PM, zthreen lists wrote:

The real question is, when will we really see this technology?

I've already seen it in action but in terms of it coming to market, maybe five years. There are a number of ways it could manifest itself. It could reside as software at the desktop level, or OS level, or even directly in the camera itself. The idea isn't to eliminate end user rendering choice, but to enhance it by starting out with a better rendering that is scene-appearance based

I put forward that digital photographers in particular, but anyone in
general, often fall prey to certain fallacies that sabotage any
attempt they would make to get accurate color.

That's not even really the goal because we know we can't get it. We see the world with a dynamic range of maybe 100,000:1 yet our camera only captures 8000:1, and print is maybe 200:1. And there is a sacrifice in color gamut as well. We can't get an exact colorimetric match. That's why images have to be rendered in order to get them to have an appearance like that of the original scene, by taking advantage of how the visual system adapts. Today's color management doesn't know anything about the adaptation of vision, or the effect of luminance.

As for projection systems that differ, there is a specification they're supposed to follow but different vendors do a better job than others. Lots of people don't know about ITU-R BT.709 but it specifies the color and tone response for HDTV and some displays do better than others still.

And for cameras that go wonky under various lighting conditions, this is hardly any different than with film where you'd use different filters for different kinds of lighting, in particular fluorescent.

Also, the RGB filters in front of a CCD or CMOS chip do not exactly mimic the human XYZ cone response function, therefore they don't see things the way we do. It was also the same with film. Two reds we see as being different, but the camera sees as the same red, or even vice versa, gave rise to the term "fugitive colors" and we still have that today. To attempt to account for the RGB filters in a camera not matching our cone response, there is at least one RGB transform matrix in a camera. It accounts for the difference between the RGB filter response of the camera, and the human XYZ cone response.

The reason why you're not getting a very good result with fluorescent is probably because there isn't a transform matrix for fluorescent in your camera and you're not using a filter for fluorescent in front of your camera lens.

The manufacturer's transform matrix applies to JPEG images from the camera. For RAW, the matrix comes from the RAW converter. In ACR, there are two, one for D65 and one for A (tungsten). You control how much of each matrix affects your image with the Temperature control. The Calibrate tab that serves to modify this RGB transform matrix.

Chris Murphy
___________________________________________________________________________

Re: A fictional case study
Posted by: "Murray DeJager"
Sun May 27, 2007 5:03 pm (PST)

zthreen lists wrote:

A quick trip over to Google tells me that green skin is sometimes
associated with multiple organ failure and if all you knew is that
my friend is dying, you might assume his skin is green.

And, lets not forget about the aliens that walk among us. I've heard that, even though their disguises are great, their true color green shows through under certain lighting conditions!

I won't say impossible, but it's only because I grew up in Fort
Worth, Texas. You see, some years back the city of Fort Worth
built a new courthouse. The architect chose to design a thoroughly
modern building, inside and out. If you've been to Texas, you know
that the old courthouses in that state are very distinctive. The
people of Fort Worth hated this new building, but the city couldn't
afford to tear down the new courthouse and rebuild in the style of
the old one, so they paid a company to erect a gigantic faux facade
that, if you don't get too close, makes the new courthouse LOOK
like a traditional courthouse made out of pink granite and marble.
But in reality, it's just a bunch of wood that's been painted pink.

I know! And I've heard this is happening all over the world!

But let's say I even tell you that I carry an 18% gray card every
where and shoot it before every series of pictures I take and go
into the menu on my 20D and load a custom white point. So the jpeg
I submitted shows a yellow man standing in front of a brown wall.
How many people would believe what they see? I've even told you
that I made sure to capture the exact white point for the scene, so
the camera didn't get confused. Who would believe their eyes? And I
don't believe there's even any software that could "correct" this
photo, as I don't know that any correction software can perform
medical diagnostic tests on the people in the picture or have any
clue that he's standing in front of brownstone (and what if bought
some of that fake stone fluffy paint from Michael's and sprayed it
on a piece of plywood, then what would the software do?)

Wow, you've got me wondering how many times I've corrected a photo and changed someones green face to a red face when in fact they were just really, really, sick!

So if you pick your colors wrong, your correction will be wrong
(but there are lots of logical...

This is a good place, IMHO, to end this "fictional" case study.

Murray DeJager
___________________________________________________________________________

Re: A fictional case study
Posted by: Gene Palmiter
Sun May 27, 2007 11:31 pm (PST)

the old one, so they paid a company to erect a gigantic faux facade
that, if you don't get too close, makes the new courthouse LOOK
like a traditional courthouse made out of pink granite and marble.
But in reality, it's just a bunch of wood that's been painted pink.

I know! And I've heard this is happening all over the world!

This is not new...faux painting dates to antiquity. Mount Vernon, too, I have been told, is made of wood painted to look like stone.

Gene Palmiter
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: "john castronovo"
Mon May 28, 2007 12:38 pm (PST)

This is the dilemma that photo labs have always been in when they print negatives. In the custom lab, it was a normal practice to make three versions to pick from and this was often just for openers. There is no such thing as an absolutely correct version from a negative and often direction from the buyer is essential or we're sunk.

One extreme example is the image of a painted box on seamless background. The lab has absolutely no idea what color either the background or the box is and the best guess involves the hope that the photographer at least matched the light to the film type. Then sometimes a photographer purposely used a color filter for effect; the lab would try to get neutral results only to discover that an off color result was expected. There are times when a great print is made from a great negative and the customer may subsequently produce an off color, maybe even stained, drug store print that he'd previously had made and actually preferred because it was "different".

With prints and scans from transparencies we can safely match the chrome for color and density, but we can also improve the image ad infinitum. But clients usually want what they've already seen or expect, and not something the lab might consider better. We can know a lot from the numbers, but we need direction from the buyer to be successful in meeting expectations.

John Castronovo
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Dan Margulis
Mon May 28, 2007 12:40 pm (PST)

Chris Murphy writes,

A prestigious professional photographer has likely been so for some
time. And arrived at that position before they had the tools and
control to get a good rendering. For years they've been used to
getting prebaked renderings from film, and now with JPEG, if that's
been their workflow.

It's only in the digital age that the cameras throw in a secret-sauce kind of modification, as you noted in your earlier post. This makes the images look better most of the time, but sometimes it makes them much harder to correct, as with our most recent examples. Most algorithms exaggerate the impact of cross-lighting, so correcting it is a much bigger deal than it was in the age of film.

Bigger factors in the historic inadequacy of Photoshop technique among photographers are that when one has a pretty good original it is easy to become too timid to take it further; that more photographers than retouchers work alone and are so less apt to study what co-workers are doing; and that many photographers learn Photoshop only from other photographers. Up until recent years, few photographers who taught digital image correction were particularly skilled at it themselves.

We should be very clear on differentiating good renderings from
artistic deviation from reality. The radically altered piece isn't
always going to get a majority of people saying "ooh yeah nice." It's
not designed to. In either case, the original photographer is the
final arbiter of what is a good rendering of the original scene,
since he was there, as well as any artistic modification taking the
image beyond that point. You can like it or not, but you can't
rationally talk about art as being right or wrong.

If you and I both think that a certain version of a photograph is poor in comparison to another, that doesn't prove that it is. If you, I, and the first 50 independent observers we can find *all* agree that the second image is better, then it *does* settle the question AFAIC--"right" and "wrong" become appropriate terms.

In a second post I talk about some of the factors that get the 100% vote.

So, let's not extrapolate beyond the singular "POS" event. What
you're characterizing here is not someone who can't identify a good
rendering when he sees it, because clearly he did recognize it. What
seems obvious to me is that this photographer simply wasn't very good
at manipulating images in Photoshop to arrive at a good rendering, a
good rendering that he identified with immediately. Perhaps he should
spend more time shooting and evaluating images, and hire someone who
is adept at translating his vision into the correct Photoshop maneuvering.

No question that with experience the number of surprises goes down. Photographers often make technical errors that they think are minor and then discover that somebody can take their final result as a starting point and make a significant, undeniable improvement in 15 seconds or less. When that happens, it's fair to say that there was something wrong with the original technique.

Somebody more experienced won't make such obvious errors, but still, it's very hard for them to know how good their result is because it's difficult to imagine what someone else might do with the same image. It happens to me, too.

Dan Margulis
___________________________________________________________________________

What makes for 100% approval?
Posted by: Dan Margulis
Mon May 28, 2007 12:40 pm (PST)

In view of recent comments about the function of color by the numbers and about viewer color preferences, a discussion of what we're trying to achieve in color correction is in order.

"Correction" means to make better. Sometimes the correction only makes things better in our own eyes and those of people who think like we do, whereas others don't like it. This type of correction is impossible to teach, as it depends on personal preference.

Many moves, however, are accepted *universally*--100% of viewers approve. Isolating these moves is what color by the numbers does. It is not always possible to spot these potential moves on a screen, which is why experienced people make such use of the numbers in the Info Palette. In terms of contrast, setting a full range as opposed to something shorter appeals to 100% of viewers. Color-wise, by the numbers is a negative, rather than affirmative, approach. It says that any number that is logically incorrect has to be changed to something that could conceivably be correct, even if that something is not the optimal value. Doing this also appeals to 100% of viewers, as opposed to leaving the impossible color there.

Prepress houses in the 1980s and early 1990s understood this, or at least their scanner operators did. Some of it (setting highlights and shadows) falls in the 100%-of-viewers-approve category. Other parts (like the mantra "clean and bright is always right") appealed to the majority of clients but not to everyone.

I first became interested in the question of 100% acceptance in, as I remember, 1993. A man who was peddling a certain acquisition method for RGB images (I can't recall whether it was called a "profile" back then) wrote an article in one of the magazines I was working for, arguing that his method was much better than any other. As evidence, he showed how it had acquired a certain image of a young boy and girl of different races, and compared it to two other common acquisition methods. The boy was wearing a white t-shirt. The author said that his own version was so clearly better than either of the other two that it proved that anyone who bought into the other methods was an idiot.

Today, we know that any profile will do well on certain images and less well on others. So, as I started to read the article, I said to myself that it was ridiculous, of course the author can show one image where his method beats the other two, but then there will be other images where it would be the worst.

Then, however, I looked closely at the three versions. One of the two competitors was clearly unacceptable, because it blew out the boy's shirt. But I preferred the other to the one that the vendor was claiming as the best. My preference was so strong that I had the magazine send me the original RGBs, so that I could see what the author was looking at, in case there had been a printing problem. There hadn't been. And yet, the author was obviously so convinced that a version I didn't like was best that he had wrapped an entire article around it.

I pulled proofs of the three images and showed them to my colleagues, asking them to state, without knowing the story, which they considered to be the best and worst of the three. Then I showed them to clients, and others, and then to classes, and to lecture groups, eventually over a thousand people. The results didn't vary with viewing condition, age, gender, or level of graphic arts experience. A bit over 50% of the audience agreed with me. A bit over 40% agreed with the author. Of these, most but not all rated the version with the blown-out white shirt as the worst of the three. However, more than five percent of the viewers, but less than ten percent, rated the version with the blown-out shirt as the *best* of the three. Upon inquiry, they always said that they understood that the shirt was a defect but that they were willing to forgive it because it made the rest of the image seem contrastier.

The author, therefore, was not necessarily wrong in saying that his version was best. The final judge of that is always the client, and based on this vote, the client might agree with him. But he was certainly wrong in his apparent assumption that the entire world would agree with him that it was the best of the three.

You would think that, having seen this error, I would not make a similar one myself, but in 1995 I did exactly that, showing two classroom versions of a picture of a baby harp seal, a white animal. The student who did the second image had been so concerned to maintain a nearly pure white that he had blasted away most of the detail. I stated that this was an example of an expert-only error and that such stupidity would be beyond the ability of mere intermediates.

Unfortunately, as a huge online response showed, many people did not agree. As with the previous author's image, I did a survey of more than a thousand people, adding a third variant that I thought no one would vote for. I was right that nobody would vote for the third one, but on the other two the vote was close to 50-50. Furthermore, it was clear from the comments that (unlike the previous image) this time, people simply could not accept the image they voted against as being even acceptable. I understood, then, why many professionals think their clients are crazy for some of the decisions they make, not realizing that half the world might make the same choice.

I then took the concept further, choosing 25 sets of three images. In each case I asked, which is the best of the three, and which the worst? And I added another significant question: granted that hundreds of people are voting, what percentage of them do you think agrees with your choice of which is the best of these three?

After getting votes from a general audience, I designated seven of the 25 sets of three for the opening day of my ACT classes. For several years thereafter (I dropped the experiment roughly in 2002) every student added their vote. One of the exercises still appears in PP5E, the Chapter 1 picture of a hog. As the text indicates, Figure 1.2C is preferred by around a 2-1 margin to Figure 1.2B.

From these three cases, you might conclude that such a large audience would never agree 100% on any image question. It's not true. Every one of the hundreds of people voting chose Figure 1.2A (which isn't all that different from 1.2B) as the worst, just as 100% of viewers chose my third version of the seal as worst. Even when people split on images there is often 100% agreement on the basics. In the seal picture there *was* 100% agreement as to which version had the better contrast, and as to which had the better color. In the hog image, the vote was very close to 100%.

In my class exercises, there aren't hundreds of voters, but usually nine, occasionally more, sometimes less. We do 20 images in the basic class, 25 in the advanced. In each case the format is single-elimination until we are down to two or, occasionally three images that are clearly better than the others, whereupon the serious comparisons begin.

This means that in the typical class there are around 300 head-to-head comparisons that are discussed by a group. Over the life of the class, I've hosted maybe 10,000 such individual comparisons of images. Most of these are 100% votes, as often people make by-the-numbers mistakes that cause the voters to reject their work immediately. Even when we have cut down to two or three images out of the original eight or nine, the vote as to which is best is still 100% more often than not. Even when it's not 100% as to which one is better overall, it's usually 100% as to the points that are considered better for one as opposed to the other. Occasionally, the vote is purely on a matter of taste. If somebody takes John's ballet picture and adds color variation and moves slightly away from blue, that would get 100% vote in favor. If there were a second version, identical for luminosity but going further away from blue, that would probably get a split vote purely as a matter of taste. But almost all the time a slight advantage in technical skill outweighs artistic preference.

Based on these votes, most of which went 100% one way or another, the following generalizations are possible.

*Failing to establish full range by setting a light point and a dark point to the extremes of what can be printed is heavily punished. It is rare to see such an image get any votes even if it is artistically superior to some other image that does set full range.

*Leaving impossible colors in the image by failing to check known areas almost always results in immediate rejection by the audience. Any version that leaves a dark-haired person's hair just slightly on the green side, for example, will lose by a 100% vote to almost any version that has no similar flaw and that uses full range.

The above two are the foundation of color by the numbers. They are not "artistic" questions. Comply, and your images compete. Don't, and they won't: the vote will be 100% against you.

Then, there are the areas in which people differ. Some of these are simple matters of taste that are difficult to generalize about: some people prefer darker, richer colors and others cleaner, brighter ones, for example. Some prefer a rather austere look and others a relatively saturated one. People's tastes notoriously vary about sharpening. And there are several areas in which preferences to move away from the art clearly exist, but the question is how far to go, and there isn't any "right" answer. These fall into two categories: the "emotional" ones that relate to how we *think* that we remember seeing things, and the "visual" ones that pertain to how humans actually see things in comparison to cameras.

The "emotional" ones:

*There is a near 100% preference for "healthy" skintones even when the original skin was sallow. This was first pointed out by research in the 1950s, but it was inaccurately described as a preference that the skin should be yellower. That's true with respect to light-skinned caucasians; it makes the skintone seem more golden. But for caucasians of medium complexion we simply prefer a more saturated color than the camera records; for dark-skinned caucasians and ethnic groups with skin at least as dark, we tend to favor adding more magenta.

*We almost always prefer greener greens than the camera records.

*Any time the sky is light, there is near 100% endorsement of the idea of making it darker (excluding the lightest parts of clouds, obviously). It is not always true that we want the sky bluer, but we almost always want it darker than in the original.

The "visual" ones:
*Human vision is self-calibrating; it neutralizes the ambient lighting. Cameras pick up all kinds of casts and mini-casts that human observers wouldn't see. As a rule, we wipe them out. However, warmish casts are tolerated to a much greater degree than coldish ones, particularly if the scene suggests warmth. Example: a swimming pool in the tropics, with palm trees. The cement lip of the pool is approximately gray in real life. Many but not all people prefer making this color redder, incorporating a cast. Some like the warming effect so much that the trees start to lose their greenness. Others prefer to stick closer to a gray. But *nobody* accepts a cool cast in this image. If the cement measures even slightly negative in the A or B channels, there is an immediate 100% vote against the image.

*Any image that is mostly of the same color, such as a portrait or a forest scene, provokes the human simultaneous-contrast mechanism. We see more color variation than the camera does. Regardless of the quality of the original photograph, a small move in this direction (often by steepening the A and B channels) gets 100% approval.

By listing all these things, I'm trying to clarify some of the confusion about "right" and "wrong" in color correction. I take the view that any move that gets near-100% approval is Right, and any that gets near-100% disapproval is Wrong. Because, as noted above, I've been through approximately 10,000 image-comparison exercises in which a group of neutral observers expressed and explained their preferences, I believe I have a much better ability to predict what will get the 100% vote than the typical person does.

When I submit my own version of an image to a class, I do not like it when 100% of the voters find that it is not as good as somebody else's. Of course, once I see that other picture, I know that they are right, but certainly I didn't know it when I finished my own. I believe most other professionals also dislike learning that 100% of observers disapprove of one of their images. The way to avoid this experience is to recognize that a few things in imaging that are truly matters of taste, but many others are absolutes. I've listed sevenfactors above. The first two are absolutes. The last five are matters of taste to the extent that people will disagree as to how far to go, but they are absolutes in the sense that it is Wrong to leave the picture the way it was found.

Dan Margulis
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Chris Murphy
Tue May 29, 2007 3:45 am (PST)

On May 28, 2007, at 1:14 PM, dmargulisnj wrote:

It's only in the digital age that the cameras throw in a secret-
sauce kind of modification, as you noted in your earlier post.

That is exactly backwards. Clearly you've never shot with Kodachrome and Fuji Velvia. They are completely different and it's specifically because of the secret sauce of the different film types that photographers had preferences between them. There is way more secret-sauce in film than a RAW file. And there's increasingly more secret sauce in JPEG than there ever was in film. There is also a lot more information about the conditions of the scene buried in the metadata in the RAW file as well, that did not exist for film.

If you and I both think that a certain version of a photograph is
poor in comparison to another, that doesn't prove that it is. If
you, I, and the first 50 independent observers we can find *all*
agree that the second image is better, then it *does* settle the
question AFAIC--"right" and "wrong" become appropriate terms.

In the case of "what looks good", that is what makes a good rendering, and that isn't subjective. It is quantifiable by having a lot of people look at various renderings and determining what most people like better. Most people will prefer a rendering that mimics the color appearance of the original scene, and that has been what's forming the basis of color appearance modeling, to take things like surrounding color, ambient light, and all sorts of non-colorimetric aspects of human vision into account. Naturally not all of these interactions are fully understood.

In the case of artistic modification, it doesn't matter if 53 people in a row don't like it. It is not right or wrong. That's the WHOLE FRIGGING POINT of art.

But you're fortunate in having used the slippery "as far as I'm concerned" (AFAIC) in qualifying when the usage of certain terms is OK for you, when anyone else would reject their appropriateness in the same context. As they say, you're entitled to your own opinions, just not your own facts. At least you've settled on this being an opinion from the outset so we're all saved a lot of time and grief.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: Potential case study
Posted by: "Murray DeJager"
Tue May 29, 2007 11:57 am (PST)

Hi Dan,

Absolutely! Prior to seeing others work with the blue ballet dancer, if someone would have shown me a that photo and asked if it could be 'corrected' to show a natural skin tone I would have said, I doubt it!

There's nothing like real world examples to show what's possible. And maybe it will give some of your C.C.'s (constant critics) the opportunity to show that their talk isn't just theory.

And even if this photo you're referring to isn't worthy, I hope someone will come up with another case study in very short order.

Murray DeJager
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: Andre Dumas
Tue May 29, 2007 11:45 am (PST)

Hello Bill,

You say: "... that is obviously not what Dan is talking about here`."

What is Dan talking about, *according to you* ? There could be different interpretations and that is not clear as far as I'm concerned (and maybe others also) please give us your views on this.

André Dumas
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: "Bill Morse"
Tue May 29, 2007 6:55 pm (PST)

Hi Andre-

Dan is not talking about artistic licence- he is "obviously," IMHO, talking about a colorimetrically "accurate" rendering.

My point, again, is that if you are going to make the green foliage magenta, for example, it sure helps to know what color green it was to begin with.

Please don't misunderstand me, I do not slavishly follow Dan's advice (I have been working in 16 bit on both color and B&W images, and am fully convinced that I get better images that way). On the other hand, when Dan or anyone else presents a wealth of information such as in his post, I would try to understand it before jumping on the author.

YMMV

Bill
--
Regards,

Bill Morse
Wm. Morse Editions
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: Matthew Rigdon
Wed May 30, 2007 3:34 am (PST)

Cameras that only use a single layer of silicon to detect color do put a filter over it, usually using a pattern called by the scientist who came up with it, Bayer. It's then up to the software in the camera or the software that processes a RAW image to determine how to mix those values together, which have already been filtered by a pattern that leaves one pixel more sensitive to red, another to blue, and two others sensitive to green (I believe). But even with the Bayer filter in place, there's still some license on what to do with the resulting data. You might decide instead of 2 to 1 to 1 to do a 3 to 1 to 1 by averaging the two greens together to make a third green value (just as an example).

The Foveon sensors use a very expensive process to lay down three layers of silicon, each of which is sensitive to one of those three prime colors and allows other frequencies to pass through to the lower layers of silicon, so at each point on the sensor you have a red, a green, and a blue value, which isn't possible with CMOS and CCD cameras that use a Bayer filter. But we've been working with Bayer cameras for so long and understand the principles so well that the additional expense of Foveon will probably never make it worthwhile.

Even with the Foveon, you have to make a decision on how to mix those colors together in order to make an image that a human can recognize. If you ever really want to get confused, try reading about color in video and wrap your head around YUV color spaces. YUV is worse than LAB.

Matthew Rigdon
http://techsonnet.blogspot.com
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: Matthew Rigdon
Wed May 30, 2007 3:38 am (PST)

Dan, first of all, great post. I've read the LAB book (although most of this discussion wouldn't have been appropriate there) but haven't finished all of the latest Professional Photoshop (I never read any of the previous editions) so I've seen some of these things scattered throughout the text, but never put together in one place. Did I miss this?

On May 28, 2007, at 10:15 AM, dmargulisnj wrote:

There is a near 100% preference for "healthy" skintones even when
the original skin was sallow. This was first pointed out by
research in the 1950s, but it was inaccurately described as a
preference that the skin should be yellower. That's true with
respect to light-skinned caucasians; it makes the skintone seem
more golden. But for caucasians of medium complexion we simply
prefer a more saturated color than the camera records; for dark-
skinned caucasians and ethnic groups with skin at least as dark, we
tend to favor adding more magenta.

Can I ask a chicken-or-the-egg type question? This sort of thing has come up in other contexts, so I'm curious. Did this research get done and then Kodak decided to formulate their film for what we now (or some of us still might call, if we remember) call the "Kodachrome" look? Or did Kodak decide to formulate their film this way and then pay for research to back it up?

I first heard some discussion about this back in the late-90's when I bought my first digital camera (a Nikon Coolpix 950). It basically came down to the Nikon versus the price-equivalent Kodak model at the time. Of course, there was no RAW mode available on either (Although the Nikon could create TIFF files, remember when that was an important feature?) so it was image processing versus image processing. There were lots of websites (still are) that uploaded multiple samples of images from both cameras that you can download and compare and the Kodak was always significantly more red-yellow than the Nikon, which tended toward blue-green.

The explanation was (and whether this ever came out of a Kodak engineer's mouth, I don't know) was that Kodak wanted to evoke a "Kodachrome" look out with the digital cameras, hence this decision. As for why Nikon went for a cooler look, this was attributed to Nikon being a Japanese company. Someone claimed that when Fuji first started out, they did the same sort of research and came to the conclusion that Japanese (and someone else claimed Europeans) preferred a cooler (blue-green) image than North Americans. Some thought that Japanese people are more sensitive to greens and blues. Who knows? I don't. But this was why Fuji film was considered too cool (in North America anyway) and explained why Fuji did better in some parts of the world than Kodak and vice versa.

I don't know where you've taught all your classes, but would you say that there's any truth to any of that or are you even in a position to make a claim either way outside of North America?

And to take this further, I'll point out that electricity's been around here for some time and that Edison invented the tungsten bulb and made it a common feature, so American's went from reading, writing, and living by candlelight (a warm colored-light) to doing the same by tungsten (which is also pretty warm). So in the fifties, people were pretty much used to seeing everything indoors through a pretty warm light. Do you think the assumptions above will hold true as we shift to a population that will probably grow up under the glow of fluorescent bulbs and, to a lesser degree, LED bulbs, probably never knowing what a lit candle looks like (I know everybody has candles on the dining room table, but do you ever light them? And do they overpower the fluorescent or LED or "pure white" tungsten lights overhead?) or do you think there will be a shift in what we psychologically think of as a healthy skintone? (CNET had a pretty good, brief article about all the lighting types http: //tinyurl.com/2dsrm6).

And just to throw another ingredient into the mix, it's far more acceptable today for mixed-race couples to have children and so there will be a tendency to far different skin tones in print and on television and film than in the past. And if you want, I wouldn't mind hearing your thoughts on what "tan-in-a-bottle" and other miracle substances are going to do what we think of as healthy skin. Hey, consider that people live longer in greater quantity and stay active, how much of an effect will that have on this emotional response to color?

Matthew Rigdon
http://techsonnet.blogspot.com
___________________________________________________________________________  

Re: Chromatic Adaptation by-the-numbers
Posted by: Matthew Rigdon
Tue May 29, 2007 6:58 pm (PST)

On May 28, 2007, at 7:15 PM, Chris Murphy wrote:

Clearly you've never shot with Kodachrome
and Fuji Velvia. They are completely different and it's specifically
because of the secret sauce of the different film types that
photographers had preferences between them. There is way more secret-
sauce in film than a RAW file. And there's increasingly more secret
sauce in JPEG than there ever was in film.

If by "secret sauce" you mean something that can't be figured out unless you work for the company, than digital cameras and RAW processors are worse about this than anybody else. With film, you were dealing with a chemical process. If you were so inclined, you could send a piece of film to a forensics lab and find out what sort of chemical base was used, what the light-sensitive particles were made of, and get an idea of how it was made. Yes, each film maker would use their own recipe, but if you shot with the same film all the time, you were able to make better educated guesses as to what would come out of your camera.

Unfortunately with any of the processors, even the RAW processors, you don't have any way of knowing what's going on inside. These companies are all really closed lipped about how the software works. It's a trade secret. And they're all different, if only so that some photographers will continue to prefer a Capture One over Adobe's ACR over Aperture over Canon's DPP, etc. The problem in the digital realm is that there's no way of knowing that a rendering will be consistent in the same frame from the shadows to the highlights. If a software engineer decides that humans are more sensitive to color in shadows than highlights, then every photo rendered by that software may have to be corrected to "fix" this decision (unless you agree with this decision, in which case you'll prefer that RAW processor over others). You don't know what the software engineer is doing because they won't talk about it. And it will affect each photo differently and in ways that can't be modeled, unlike with film which was still subject to the laws of chemistry which are pretty rigid and well-known.

And of course, the software engineer could have always made a mistake that slipped through QA.

Of course, the software will always have "secret sauce" because it's what keeps customers around. If all RAW processors worked the same, no one would buy anything but Photoshop and use ACR, maybe a few people would buy Lightroom if they needed organizational capabilities.

Matthew Rigdon
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Mark Segal
Wed May 30, 2007 3:37 am (PST)

There's nothing unfortunate because we don't need to know. That's one of the advantages of raw processing. We get relatively unadulturated data from the camera into the program, and we have vast amounts of control over its rendering just by making adjustments and looking at the results on your monitor. Seems very controllable and transparent to me - much, much more so than I could have ever had with film - indeed one of the reasons I switched to digital capture. For example, we make our own curves and see the results as we do it - in a properly color-managed workflow. No way you could change the characteristic curve of a roll of film.

Mark Segal
___________________________________________________________________________


Re: Chromatic Adaptation by-the-numbers
Posted by: Dan Margulis
Wed May 30, 2007 8:46 am (PST)

Chris Murphy writes,

That is exactly backwards. Clearly you've never shot with Kodachrome
and Fuji Velvia.

I admit my ignorance of these films' characteristics--I doubt that I've scanned more than 10,000 originals of either type. So you will have to forgive me for not being aware of the features that you seem to be referring to.

They are completely different and it's specifically
because of the secret sauce of the different film types that
photographers had preferences between them.

"Secret sauce" implies something secret. At the time I was doing this scanning there was no secret about how these films behaved. It was so non-secret, in fact, that a folk-rock singer who AFAIK had no graphic arts experience wrote a top-of-the-charts song extolling the behavior of one of them. By contrast, what acquisition modules and JPEG routines do to files are a big secret to almost everyone.

I was under the impression that the character of the image had nothing to do with the way those two types of film behaved. For example, I had thought that if you take two identically-exposed pictures of the same object in the same lighting, it would appear as the same color on the film regardless of what other objects appeared in the image or how the photographer framed the shot. Obviously the film must not behave this way, because otherwise your analogy would make no sense. The defining feature of the secret sauce of automated rendition is that irrelevant contents of the image can change known colors. That's why we have so many issues today with crossing casts, a problem that was not nearly as prevalent in the age of film.

Similarly, I didn't know that these films suppressed highlights and shadows, the way every digital acquisition algorithm does if left to its own devices. I had supposed it possible to manufacture film that was insensitive to relatively bright light, but thought it would be difficult to make it simultaneously insensitive to relative darkness. Yet it must have happened, because otherwise your analogy would make no sense. Today, we've got to use the Shadow/Highlight command, or a raw acquisition module, or inverted overlay blending, to enhance the extremes. This was rare in the days of film.

Also, I did not know that these films were capable of making an independent decision as to what was white. I had foolishly supposed that if the light hitting it wasn't neutral, the film would record as non-neutral. But apparently you are saying that the film itself makes an effort to whiten it--otherwise your analogy would make no sense. So often today we find images with neutrally correct highlights and competing casts throughout the rest of the picture--a more difficult assignment than back in film days, where we'd get one uniform cast.

Further, I didn't realize that the film attempted to lighten the image when it thought that the photographer had underexposed it, darken it when it thought the image was overexposed, extend range of interest areas, identify fleshtones, and discard image information that the film decided was irrelevant. But apparently the film must have done all those things, or your analogy would make no sense.

Personally, I think you may be confusing the film with the scanner operator, who *did* do all those secret-sauce kind of hijinks, hopefully in a more intelligent way. The two were actually rather easily distinguished: the film was the thin stuff with the plastic coating on one side and the scanner operator's fingerprints on the other. The scanner operator was the one with a cigarette in one hand, a drink in the other, and a propensity to make rude remarks about photographers and calibrationists.

There is way more secret-sauce in film than a RAW file.

Right up until the time you open said file and start messing with it, whereupon you get all the secret sauce that the module's designers built in.

In the case of "what looks good", that is what makes a good
rendering, and that isn't subjective. It is quantifiable by having a
lot of people look at various renderings and determining what most
people like better. Most people will prefer a rendering that mimics
the color appearance of the original scene, and that has been what's
forming the basis of color appearance modeling, to take things like
surrounding color, ambient light, and all sorts of non-colorimetric
aspects of human vision into account. Naturally not all of these
interactions are fully understood.

This is a good summary that is worth re-reading.

In the case of artistic modification, it doesn't matter if 53 people
in a row don't like it. It is not right or wrong. That's the WHOLE
FRIGGING POINT of art.

Yes, but you've forgotten one thing. If the vote is 100% against it, that means the artist doesn't like it either.

But you're fortunate in having used the slippery "as far as I'm
concerned" (AFAIC) in qualifying when the usage of certain terms is
OK for you, when anyone else would reject their appropriateness in
the same context. As they say, you're entitled to your own opinions,
just not your own facts. At least you've settled on this being an
opinion from the outset so we're all saved a lot of time and grief.

I'm not sure that too many graphic artists of any discipline would reject my contention that if 100% of observers, including the artist, dislike a certain rendition there is something wrong with it. But if not, it does open a great new revenue source for many of us. We take some image that we feel is no good, verify that everybody else also thinks it's no good, and give it to the client, explaining that it's art.

Dan Margulis
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Gene Palmiter
Wed May 30, 2007 12:58 pm (PST)

Dan Margulis said:

I'm not sure that too many graphic artists of any discipline would reject
my contention that if 100% of observers, including the artist, dislike a
certain rendition there is something wrong with it. But if not, it does open
a great new revenue source for many of us. We take some image that we feel
is no good, verify that everybody else also thinks it's no good, and give it
to the client, explaining that it's art.

But, aren't we almost there? Young minds of mush go to college and learn what 'art' is and then get hired by corporations to buy 'art'. Does anybody really like most of the 'art' that is on public display these days? Reviews of same are mostly curiculum vita of the 'artists' and the price paid. There are lots of emperors with new clothes these days.

But, we are not really talking 'art' here, we are making images that will reproduce well using a particular technology and that will grab the attention of the viewer long enough to impart some sort of feeling or message. We are communicators and art that nobody likes is just the artist mumbling to themselves...a sign of probable insanity.

Gene Palmiter
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: "Ron Kelly"
Wed May 30, 2007 7:52 am (PST)

Matthew:

I'm not so sure you're right about that; a higher quality sensor will always have a market.

I have wondered if the Foveon sensor would predominate with time, which hasn't happened. It's definitely superior to the Bayer pattern sensors, on a pixel-for-pixel basis. I guess it has other problems which keep it from being more widely adopted.

Ron Kelly
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: palmiter_gene
Wed May 30, 2007 12:56 pm (PST)

Last I heard, about a year ago, there seemed to be a real advantage to the Fuji sensor that has large and small sensor elements that provide a wider dynamic range. This sensor is at an angle that provides a sharper image than an equal number of pixels from a normal Bayern pattern. I would like to see what combinations of these technologies with the Foveon might accomplish but as all are proprietary it will be some time until this can be tried.

Right now, the improvements seem to be coming from putting more resolution (pixels) on the chip. Is anybody claiming to achieve a better image than the Hassleblad with 39mp? I doubt it...but time will tell.

Gene Palmiter
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: "Hoffner, Randall N"
Wed May 30, 2007 12:57 pm (PST)

One of the characteristics of color film is that, because the R, G, and B - sensitive portions are layered, there is a decreasing sensitivity as the light passes through each layer. Red, for example, is on the bottom layer, next to the substrate, and film is less sensitive to reds than to blues and greens, because some of the red light is filtered out before it ever reaches the red-sensitive layer. Green is on top, and guess what? Film is most sensitive to green light. This is felt to be the cause of the frequent complaint by cinematographers that video, which is typically equally sensitive to R, G, and B, does a bad job of rendering skin tones compared to film, because it makes skin look too red. I presume that the Foveon sensor might be expected to produce a similar result to film in this respect.

And, I never really thought about it before, but at a conference in February, someone who was talking about the newest bleeding-edge video cameras that can shoot 4K and the like (raw files, in some cases, as a matter of fact) - contrary to the historical situation, all the highest-end video cameras that are showing up today use Bayer-Pattern sensors, not separate RGB sensors and prisms, etc.

Randy Hoffner
___________________________________________________________________________
.
Re: What makes for 100% approval?
Posted by: Andre Dumas
Wed May 30, 2007 1:00 pm (PST)

Hello Bill,

"Colorimetrically "accurate" rendering" is a subject that I'm interested in. How to define that term, I mean: when are my "corrected" photograph colorimetrically accurate ? When I "correct" an image when does it become no longer colorimetrically accurate ?

If the BTN rule on rendering is to mean "colorimetrically accurate" then we must have an accurate definition of that term relative to BTN.

When I prepare an image from my own photographs I start with a raw adjusted to look exactly like the original scene, I don't consider artistic moves at that stage, just like when I do a painting in nature. Then I transfer the image into Photoshop and begin my interpretation and if necessary I will go back to the raw and restart the process.

My end product will never be like the original scene, it may be close but almost never like the original. Can this be called "colorimetrically accurate or inaccurate?

Since I do a lot of landscapes, I am particularly finicky about greens. I began learning about them with Dan and now I have progress beyond Dan's definition of "greens by the number". Not that Dan is wrong but rather that BTN needs to be taken a few steps further regarding the color of things in particular and chromatic adaptation in general.

André
___________________________________________________________________________
.
Re: Potential case study
Posted by: Mark Segal
Wed May 30, 2007 12:57 pm (PST)

Dan,

This is out of date. The Raw Converter in CS3 (ACR-4) is far more powerful and offers many new and improved options for correcting and modifying colour balance, hue, brightness and saturation of individual colours, and overall image brightness and contrast whether in colour or greyscale.

Mark Segal
___________________________________________________________________________

Re: Potential case study
Posted by: Mark Segal
Wed May 30, 2007 12:58 pm (PST)

If you are talking about ACR-3, it is not the same program as ACR-4. ACR-4 has more capability and MAY be able to deal successfully with major issues in that image - I don't have the raw file to test it -- one needs the raw file for this. Saying this, I'm not suggesting ALL the work can be done in a raw converter - it's not the intention of that application, nor of Lightroom. But the more that can be done before rendering the image into a three or four channel image the better. Aperture is only available on the Mac platform and therefore automatically excludes half the imaging market, and automatically excludes me from the discussion of Aperture, except that several folks I know who do have Aperture, Lightroom and ACR don't think Aperture brings anything better to the table than we have in the others. I don't see the point of comparing drum-scan technology of the 1980s with what can be done using today's hardware and software. It's for sure worlds apart in terms of workflow efficiency and flexibility, and most likely now distanced in respect of achievable quality.

Mark Segal

___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: "Bill Morse"
Wed May 30, 2007 3:55 pm (PST)

Hi Andre-

Yes, the process you decribe is almost identical to what I described as learning the rules, then breaking them. We all do it, to the extent that we take artistic licence, and/or work towards abstraction as opposed to realism. For me, Dan's post was just another guide-post (pardon the pun) on the way to understanding what "other people" (maybe even ~100%?) percieve as reality. From there, we can either give them what they expect (best for paid gigs?), or not.

Regards,

Bill Morse
Wm. Morse Editions
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: "john castronovo”
Wed May 30, 2007 8:55 pm (PST)

Right now, the improvements seem to be coming from putting more resolution
(pixels) on the chip. Is anybody claiming to achieve a better image than the
Hassleblad with 39mp? I doubt it...but time will tell.

Sure. Have a look at the following comparison done by the folks at Betterlight who's scan back makes the Hasse (and other Bayer matrix based cameras) look pretty pathetic. This clearly shows the pitfalls in both resolution and color accuracy associated with today's sensor technology versus a true RGB scan.
http://www.betterlight.com/compare_DSLR.html

John Castronovo
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: "Les De Moss"
Wed May 30, 2007 6:03 pm (PST)

John W Lund writes:

Yes, Kodachrome does it's version of the apple's red color, so does Velvia, generally
independent of other objects in the scene (unless they reflect some light back onto the
apple, to be sure).

Right. And in that case, the object (apple) has been changed.

I don't see this as inherent to digital capture, since 3rd-party raw converters do not follow
this behavior.

More accurate to say it's inherent in processing digital images via "smart" algorithms... which for consumer digicams and their users, do a remarkably good job on a high percentage of images. It's the unwanted, forced trade-offs that occur in automatic processing that cause pro's to use alternative processing methods such as Raw conversion.

Les De Moss
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Matthew Rigdon
Wed May 30, 2007 8:53 pm (PST)

I thought what Dan was claiming is that if you shot a red apple on a white background with Kodachrome, then shot the same red apple on a yellow background with Kodachrome, you'd still see the exact same red apple. In fact, taking chromatic adaptation out of the picture, if you cut the apple out of both pictures and put them side by side, they'd look exactly the same.

Now, this assumes that the guy at the lab developed the film and the print by-the-numbers, i.e., he took the temperature of each of the chemicals, he went to the data book and looked up the exact time to leave the film or print in, and he used an accurate stopwatch (note accurate, not necessarily consistent, but let's hope the watch is also consistent).

The same thing happens with RAW processors. RAW processors from different companies have different output. Part of this is so you'll buy their product. If they all looked the same, no one would buy anything but Photoshop and use ACR. Why pay extra if it doesn't look different? But I've seldom noticed that ACR's automatically determined white point temperature matched up with what my camera determined to be the white point temperature at the scene. Sometimes they differ wildly; sometimes it's just a few hundred K, but figure in a few hundred K here and there, plus some differences in gamma and color space, and suddenly things turn out a lot different.

Even if you know the color space and the gamma, though, you still can't predict what a RAW processor will do to your image because you can't predict what it's going to pick as a white point temperature. And without that white point temperature, you have no way to make color. And who knows whether someone on the programming team decided "everything's too cool" and put in a red-yellow boost to "correct" it.

I saw plenty of examples this weekend of how the auto white balancing in professional video cameras can be absolutely destroyed by the introduction of a colored light, not on the subject being shot, but on a white wall behind them. Cameras don't see anything like what people see. And all of these automatic sensors are just algorithms, each written by a different software engineer at a different company working off a different set of hypotheses about how the human visual system works.

Matthew Rigdon
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Andrew S. Webb"
Wed May 30, 2007 6:04 pm (PST)

On May 4, 2007, at 9:17 AM, dmargulisnj arranged some pixels so they looked like this:

No, I'm right. I tested the module carefully in preparation for PP5E. This
feature particularly, because I was dubious that the implementation would
have been done in the way it was.

Does this cover the capabilities of Camera RAW 4, or are you just talking about the version that shipped with CS2? Not at al the same thing. I'm still investigating both versions to try to see for myself. How did you test the module for PP5E?

Cheers,

/andrew webb
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: "John William Lund"
Thu May 31, 2007 1:37 pm (PST)

--- Matthew Rigdon wrote:

I thought what Dan was claiming is that if you shot a red apple on a
white background with Kodachrome, then shot the same red apple on a
yellow background with Kodachrome, you'd still see the exact same red
apple. In fact, taking chromatic adaptation out of the picture, if
you cut the apple out of both pictures and put them side by side,
they'd look exactly the same.

...<snip>...

The same thing happens with RAW processors. RAW processors from
different companies have different output. Part of this is so you'll
buy their product. If they all looked the same, no one would buy
anything but Photoshop and use ACR. Why pay extra if it doesn't look
different? But I've seldom noticed that ACR's automatically
determined white point temperature matched up with what my camera
determined to be the white point temperature at the scene.

-- Hello Matthew,

Yes, I'm completely with you - until you get to that last sentence.

Why would anyone serious about color reproduction ever use auto white point? Whether set by the camera or by a raw converter, I would never expect "accurate" results that way - at best, maybe a quick way to get a starting point.

Using any "automatically" determined white point technique seems like accepting a camera's auto exposure recommendation. If accuracy is important, you get better results reading a carefully placed Macbeth card, and metering appropriate scene values, respectively.

This whole line of portraying all digital capture as being riddled with uncontrollable "secret sauce", then holding up in-camera jPeg processing, auto white point algorythyms, etc. as examples - seems like a bogus straw-man argument to me. Sure, one can accept Canon's idea of what a 'pretty picture' should be, just like one could accept a lab's machine prints from negatives, but neither is a fair representation of the reproduction capabilities of the respective processes.

(Phew - sorry about that last tortured sentence)

Regards,

John W Lund
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Chris Murphy
Thu May 31, 2007 1:37 pm (PST)

Matt, you're confused. Accurate is not how a human sees. We have all kinds of visual effects that make us think we are seeing things that we are in fact not seeing. It's part of the visual experience. Hence why there are terms like colorimetry, which is about color accuracy and is based on how humans see in very specific lighting conditions, and working with uniform colors. All of those tests were done with light samples, not images.

Color appearance is about how we see the world. It's affect by ambient light, luminance, color temperature, all sorts of visual effects, etc. And this is the realm of color appearance models that predict quite, but not all, of what we know about seeing the real world around us.

It is possible, today, to produce a scene-referred (read colorimetric match to the scene) from a raw capture. You really cannot do this with JPEG or with film. And if you look at that colorimetric match, changes are you won't like it compared to your memory of the scene because the viewing conditions have radically changed. Colorimetry wasn't designed with this in mind. Color appearance modeling is.

Our attempts to get that scene to print match is all about rendering. What compromises need to be made in order to get that great, vivid, contrasty scene to look good in print. That's rendering. And color management has a role to play in both, because ICC v4 does allow for both paths. Arguably we need more intelligent CMMs to do some fancier handling of rendering than a prebaked perceptual intent in an ICC profile can really perform, but in principle color management can do accurate as well as inaccurate but pleasing rendering.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Chris Murphy
Thu May 31, 2007 1:38 pm (PST)

On May 29, 2007, at 9:11 PM, zthreen lists wrote:

If by "secret sauce" you mean something that can't be figured out
unless you work for the company, than digital cameras and RAW
processors are worse about this than anybody else.

a.) No I do not mean something that can't be figured out unless you work for the company.

b.) What I mean by secret-sauce, is that you aren't getting a flat rendering with *any* film type. It's not possible. They all have built-in proprietary rendering in their chemistry.

c.) There is proprietary, built-in, rendering (a "look") for camera JPEG.

d.) There is very little, in some cases no, built-in rendering for camera raw.

With film, you
were dealing with a chemical process. If you were so inclined, you
could send a piece of film to a forensics lab and find out what sort
of chemical base was used, what the light-sensitive particles were
made of, and get an idea of how it was made.

If you're so inclined, you can use a monochronometer and determine the spectral sensitivities of the camera, and built multiple illuminant specific camera characterization matrices. If you're so inclined you can figure out what's in a raw file. People are doing these things.

Yes, each film maker
would use their own recipe, but if you shot with the same film all
the time, you were able to make better educated guesses as to what
would come out of your camera.

This is vastly easier with raw, because the capture is flat, the raw data is largely untouched, especially in comparison to film or JPEG. Raw, in every camera so far, has not been rendered for output.

Unfortunately with any of the processors, even the RAW processors,
you don't have any way of knowing what's going on inside.

It's true we don't know the specifics of what's going on, but based on the behavior of a lot of different cameras, and the reverse engineering of the raw file format from these cameras, we do know that if they're massaged values, they haven't been massaged a lot. i.e. we know that raw is unrendered. It's not nearly as proprietary as either film or JPEG.

Thesecompanies are all really closed lipped about how the software works.
It's a trade secret. And they're all different, if only so that some
photographers will continue to prefer a Capture One over Adobe's ACR
over Aperture over Canon's DPP, etc. The problem in the digital realm
is that there's no way of knowing that a rendering will be consistent
in the same frame from the shadows to the highlights. If a software
engineer decides that humans are more sensitive to color in shadows
than highlights, then every photo rendered by that software may have
to be corrected to "fix" this decision (unless you agree with this
decision, in which case you'll prefer that RAW processor over
others). You don't know what the software engineer is doing because
they won't talk about it. And it will affect each photo differently
and in ways that can't be modeled, unlike with film which was still
subject to the laws of chemistry which are pretty rigid and well-known.

Yeah and some people didn't trust a photo lab to develop and print their pictures either, because that whole process was like a black box, so they developed their own chemistries. In a sense, this seems much more complex and hidden today because the skill set involves programming rather than chemistry. But from a big picture view we have about as many barriers as before, but we definitely have a lot more information in a raw file than we did with film, and we also have unrendered files. That means the control of rendering is in the hands of users. You are certainly welcome to use any number of raw processing engines out there with various presets that will give your raw image a default rendering that the manufacturer of that engine prefers (or thinks you will prefer). If you don't like it, it's vasty, vastly easier to create your own rendering today without needing to program or deal with film chemistry.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: Andre Dumas
Thu May 31, 2007 7:30 am (PST)

Hello Dan,

I have read your message and I do understand your point of view. Mine is that when I render an image I am making an artistic statement and, I realize now, that it cannot be bound by the specificities of your BTN function.

I just finished reading Annie Leibovitz's "A Photographer's life 1990 – 2005", not surprisingly, none of her full color photographs would comply with your definition of by the number, it's a question of style, Annie Leibovitz has her artistic style, you have yours which you call BTN, I have mine, etc…, and that is the way it should be.

André Dumas
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: Matthew Rigdon
Thu May 31, 2007 7:31 am (PST)

Up until Canon spent a boatload of money developing their DIGIC sensors, no one believed that you could make a professional quality image sensor using CMOS technology. CMOS is the basis for most CPU manufacturing, so the chips are relatively cheap to produce. Canon decided to finally take on the initial investment cost of making a set of electronic brains that could decipher a good image from a CMOS chip in order to get to the other side, where the lower cost of production would kick in.

This is why I don't think Foveon will ever take off. It just costs too much and it always will. For one, you can't use the same silicon wafers that everyone else in the industry does. You have to have special silicon wafers manufactured that will allow different frequencies of light to pass through. And who needs that besides camera makers? On top of that, you have to put those three layers on in separate steps, another level of cost (although other designs also require this, so it's not as big of a hit). And you're also using a system that requires a whole different set of logic, one that isn't based on processing voltages that have passed through a Bayer filter. Because even before you save out a RAW file, there has to be some processing done to all those voltage levels to turn them into digital bits. And as pointed out, no matter how hard you try, you won't be able to pass red frequencies through two layers of silicon perfectly (if red is the bottom layer, I don't know what the order is).

I wouldn't be surprised to see CCDs disappear over the next couple of decades. The amount of engineering that's devoted to CMOS is staggering because the potential pay-off is huge over the long run. At the conference, they were probably drooling over the new RED camera, which has the potential to completely change the high-end HD camera business (potential, I didn't say it would, they're taking on some big players with deep pockets).

As for Fuji's design, it's always gotten raves for it's superior dynamic range, but it's always been hurt by cost issues. The first camera they built used a Nikon N80 body (a cheaper prosumer film body) with their sensor, but it cost somewhere around $3000. For that price, people wanted a lot more in the physical body. The more recent versions have gotten better on that level, but moving to CMOS allowed Canon to produce full-frame sensors at a price that won't break the bank, making it far more easier for film shooters to transition over since you don't have to re-learn how to frame your shots. I think most photographers, given the choice between superior dynamic range or knowing that a 50mm lens will still look like a 50mm lens, will choose the latter. Partly because they've gotten used to shooting on film and losing dynamic range to a scanner somewhere along the way all this time.

Matthew Rigdon
___________________________________________________________________________

Re: Chromatic Adaptation by-the-numbers
Posted by: Matthew Rigdon
Fri Jun 1, 2007 5:07 am (PST)

On May 31, 2007, at 11:07 AM, Chris Murphy wrote:

Matt, you're confused. Accurate is not how a human sees. We have all
kinds of visual effects that make us think we are seeing things that
we are in fact not seeing. It's part of the visual experience. Hence
why there are terms like colorimetry, which is about color accuracy
and is based on how humans see in very specific lighting conditions,
and working with uniform colors. All of those tests were done with
light samples, not images.

No, I see that you're confused because you never work with humans. If I'm standing in a room watching someone on stage with a purple background and the camera takes a picture of someone of a completely different color standing in front of a blue background, then the camera isn't accurate. I may have needed to override auto white balance on the camera to fix it, but that requires human intervention. And until cameras can see the way humans do, we can hope for consistency, but not always accuracy.

And if everyone in the room sees a purple background, it doesn't matter what the camera says. the human visual is the scale. It may be a sliding scale, but until the machines take over and they pay the bills, what the humans want is all that counts.

And some of us can't walk around and hold a card in front of everything we shoot. It's just not practical.

Matthew Rigdon
___________________________________________________________________________

Re: Potential case study
Posted by: "Rick Gordon"
Sat Jun 2, 2007 3:03 pm (PST)

Some interesting variants on this theme can be played by:

1) Instead of using a stamped (Merged Visible) copy of the image for the contrast layer, use whatever channel gives the most bang for the buck or create a special grayscale variant for the purpose using Calculations, Channel Blending, of Photoshop CS3's awesome Black & White filter.

2) Try different overlay mode variants (soft light, hard light, vivid light, etc.), and adjust sliders and opacities with utter disregard to any supposed correct value. They will be all over the map.

3) Try doing it in a different sequence. I most often:

a) Create the gray channel.

b) View it in Normal mode as I set the High-Pass value.

c) Change to an overlay mode -- experiment on which one to choose.

d) set Sliders.

e) Set opacity.

4) As added possibilities to the above, consider:

a) Creating the whole setup using Photoshop CS3's Smart filters, along with layer clipping masks to be able to tweak all elements of the setup visually and interactively, rather than making sequential choices. You can always flatten it after you're satisfied.

b) Once you've got it, Merge Visible and the hide or dump the setup layers, duplicate the stamped layer, set one of them to Darken and the other the Lighten, and then adjust their opacities separately. If you take this route, it's useful not to set the opacity of the overlay layer too low before doing the Merge Visible, so you have something to back off from.

As Andrew says, this basic approach is fantastic, and there are some other different but related approaches that can also work midtone magic. (But not right now ;)

Rick Gordon
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Mon Jun 4, 2007 9:31 am (PST)

Andrew Webb writes,

Does this cover the capabilities of Camera RAW 4, or are you just
talking about the version that shipped with CS2? Not at al the same
thing. I'm still investigating both versions to try to see for
myself. How did you test the module for PP5E?

The book covers CS2. Readers are cautioned both at the beginning and end of the chapter, "Raw modules are currently the hottest area of imaging software development. It should be anticipated that capabilities and speed will improve rapidly over the coming years."

The testing methodology is outlined in Chapter 16 of the book. The list played an important role. Since the previous edition had come out in 2002, I wanted to replace nearly all the images with ones more typical of today's professional practice. I therefore asked photographers on this list for their support and they responded by graciously volunteering to send more images than I could possibly look at. I had to decline most offers with thanks. I asked for files from people who would provide the broadest diversity of shooting conditions and camera models. The total quantity of volunteered images that were reviewed for PP5E was around 80 gigabytes. Those that were submitted in raw format came from ten different models of five different manufacturers, several thousand individual files.

I first examined the files that were in Raw+JPEG format. I batch-opened all the raw files using Camera Raw's auto adjustments to see how the artificial intelligences compared. A Camera Raw auto adjustment never gets identical color to that of the camera manufacturer's JPEG. However, in more than half of cases the two methods produced similar results. In the cases where there was a significant difference, the cameras' JPEGs were superior to the Camera Raw auto adjustments in roughly a 3-1 ratio. The book shows side-by-sides JPEG vs. Camera Raw auto adjustment of two images, one in which Camera Raw did much better than the camera and once much worse.

I did not break these results down by camera manufacturer because I think few people are interested in using Camera Raw as an auto-adjustment tool--if so, they'd be better off just opening the JPEG.

Next, I looked at the JPEGs versus examining and adjusting in Camera Raw. The purpose was not to make final corrections but to see how frequently the JPEGging had thrown away or damaged important image information that Camera Raw might be able to recover. The conclusion was that in 90+% of cases the JPEGs were perfectly sound and could be corrected into files of at least as high quality (excluding any artifacts of compression) as if they had been acquired by any means in Camera Raw and then corrected either there or in Photoshop proper.

Of the remainder, in some cases enough highlight detail had been lost during the JPEG that the file could no longer be corrected to the same quality as if it had been acquired sensibly in Camera Raw and then corrected. Some others *could* have been corrected starting with the JPEG but it would have been a lot more work than starting in Camera Raw. I did not distinguish between these two types of problems except to be able to acknowledge that in at least some cases JPEG leaves us with a problem that has no solution.

Moving away from JPEGs into the raw files proper, I then examined each Camera Raw command to see its functions. With respect to White Balance, Brightness, XXX, and curves, I also scripted a routine to batch open multiple variants and spreadsheet the results, to reveal the exact mathematics behind the command. This was necessary because, as noted in the post you responded to, I simply could not believe that a major vendor was using a method this crude to open range.

The white balance routine, by contrast, is well thought out and original AFAIK. In CS2, however, I think a lot of people would have trouble figuring out how to use it properly. I am glad to see that it has been made more self-explanatory in CS3. But the range-opening routines are inferior, and range opening is the most fundamental operation of the module, the one function that can't be omitted in most images. All the add-ons in the world can't compensate for improper ranges.

Also, that it was done this way suggests a misunderstanding of what the module's desired function is. The range-opening commands use the same master-channel approach that a JPEGging algorithm does. That's right for most images (otherwise the vendors wouldn't use it at all) but the very fact that we are trying to manipulate the image in the raw module indicates that we probably don't want to use that approach.

I studied whether there were any circumstances under which image-by-image manipulations in Camera Raw followed (if necessary) by manipulations in Photoshop would ever yield superior quality to a conservative, damage-free Camera Raw acquisition followed by Photoshop manipulation. After testing around 100 images, I concluded that in some cases opening the range in Camera Raw actually damaged the image to the point that it was no longer possible to get a good result without excessive effort, if it was possible at all. In most cases the impact was small. There were no cases I could identify where one could get better image-by-image results by using any of Camera Raw's functionality.

I also tested around a dozen images where the capture had serious exposure defects, as I had thought that recovering from such difficulties might be a strength of the module. Most of the time such pictures aren't very important. At such times Camera Raw offers an attractive way to get close to the desired result quickly. However, if time is not a factor, better results are often achievable by acquiring conservatively and then assigning a false profile.

In the chapter and the enclosed CD, I showed full exercises of three separate raw images illustrating these points.

1) In an image of a delicately-detailed poppy, I showed how detail could be enhanced if the image was acquired conservatively in Camera Raw, but that if we had established the desired range using its curve or Brightness command,(or, presumably, if we had had a JPEG) it would no longer have been possible to achieve equivalent results.

2) An image of an elderly couple shot indoors showed a yellow cast. After a conservative, flat-looking acquisition in Camera Raw, the image can be corrected to my satisfaction with one set of curves. I showed what happened if an attempt had made to correct the problem with Camera Raw's range-opening and/or White Balance routines. Doing so introduced the common digital defect of conflicting casts at different darknesses. Readers were referred to the correction of a similar yellow-cast, elderly woman image in another chapter. As only a JPEG was available for that image, it resembled the full-range acquisition in Camera Raw, and was difficult to fix.

3) A badly underexposed wildlife picture had (when the range was finally opened) critical details in both highlights and shadows. I suggested that such images are best corrected by creating two independent versions and then merging them. I showed how to generate the two versions in Camera Raw and how to further correct and merge them in Photoshop. There would be other ways to get the same result, or possibly better, but not as quickly IMHO.

When the chapter draft was completed, it was carefully reviewed by ten beta readers who had been selected for their diversity being representative of the readership as a whole. The skill level ranged from intermediate to expert. All had access to the raw files and most redid the exercises to verify the results. This draft was one of the two highest-rated chapters in PP5E.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Andrew S. Webb"
Mon Jun 4, 2007 11:51 am (PST)

Thanks, Dan, I appreciate the time you took to explain all of that. I have PP5E, but don't remember this part. I've been reading your books from the beginning, so I tend to pick and choose a bit in the later editions. I'll go back and read that chapter and look at the files.

With regard to the master-curve-type adjustment, I understand what you're saying now.

Cheers,

_andrew webb
___________________________________________________________________________

Re: Digest Number 2434
Posted by: Jeff Schewe
Tue Jun 5, 2007 7:41 pm (PST)

On 6/5/07 7:18 AM, "Dan" wrote:

The book covers CS2. Readers are cautioned both at the beginning and end
of the chapter, "Raw modules are currently the hottest area of imaging
software development. It should be anticipated that capabilities and speed
will improve rapidly over the coming years."

That was then...this is now...

I then examined each Camera Raw command to see its functions.

Camera Raw 4.x pretty much blows the test out of the water since between Camera Raw 3.6 and 3.7, CR evolved to be able to render Lightroom control settings and in Camera Raw 4, all the controls now reflect the Lightroom version of the Camera Raw processing pipeline and the entire logic of the auto settings changed. So, other than WB and Calibrate, it's all different now...

But the range-opening routines are inferior, and range opening is the
most fundamental operation of the module, the one function that can't be
omitted in most images.

So, are you referring to Highlight Recovery and Fill Light? I'm not familiar with the term "range-opening"...did you make that up Dan?

The range-opening commands use the same master-channel approach that a
JPEGging algorithm does. That's right for most images (otherwise the vendors
wouldn't use it at all) but the very fact that we are trying to manipulate the
image in the raw module indicates that we probably don't want to use that
approach.

So, what's this "master-channel" thingie you're talking about. Is this another coined "catch-phrase"? Or, are you referring to the fact that tone adjustments are being done on the luminance data? Since the original raw data is in a more or less linear gamma, tone mapping before the eventual gamma encoded color space transform is pretty par for the course ya know?

You wouldn't be suggesting that the tone mapping of luminance data should wait until AFTER it's been gamma encoded-would you? I mean, you know enough about curves in Lab to know that mapping the luminance data without the impact of color data is useful...right?

This draft was one of the two highest-rated chapters in PP5E.

Plan on updating the chapter to reflect the current processing scheme? Cause that was then and this is now...ya know?

Regards,
Jeff Schewe
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Rick Gordon"
Wed Jun 6, 2007 4:12 am (PST)

We probably should wait for Dan to weigh in on what he meant as "range opening", but my guess is that he is referring to the idea of expansion (or in the case of Camera Raw, also possibly contraction) of the luminosity range to cover (in typical images, but not all images) a shadow point just distinguishable from pure black and a highlight point just distinguishable from pure white, plus the possibility of very limited points of pure black or pure white.

In Camera Raw terms, that probably relates to setting the highlight with the Exposure value and the shadow with the Black value.

Though I am an avid Adobe Camera Raw user and supporter, I might guess that part of Dan's complaint is that part of the resulting effect of wide expansion the luminosity range (as might be required in a scene in low light) -- particularly when there are big positive moves made with Contrast, Fill Light, or Curves -- is an increase in saturation, which must then be countered with various other moves to reduce the saturation back to a desirable range. It may be all doable, but sometimes the pure luminosity moves of an L curve or a luminosity blending mode, might realistically be wished for.

(Now I can wait and see if I've completely misinterpreted Dan's intent.)

Rick Gordon

___________________________________________________

RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Tue Jun 5, 2007 11:46 pm (PST)

Well, I'm very red-faced....seems as though I _THOUGHT_ I had changed the subject line from 'Digest' to the current topic thread: 'Color theory and practice' but I obviously failed to do so...my apologies to the list and to Dan for my oversight...

But I am curious about the new terminology that Dan is developing...let me throw some others out that are relatively new...

'Parametric Editing' which is the editing of parameters vs. pixel editing like in Photoshop. The new parametric editing term is something Mark Hamburg coined to explain how image editing with Lightroom (and Camera Raw since it shares the same pipeline and parameters now) is different than pixel editing in like, Photoshop.

Thomas Knoll also cooked a new term-"just in time image processing" meaning the actual processing of a raw file (and now jpgs & tiffs since LR/CR now does both) just in time for the intended use...particularly relates to Lightroom since raw images can be stored with the parametric edits and processed just in time for the intended use...but also refers to parametric edits done in Camera Raw & Bridge.

So, those are a couple of new terms to throw into the mix.

:~)

Regards,
Jeff Schewe
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Lee Clawson"
Wed Jun 6, 2007 8:17 am (PST)

Dan,

I haven't heard "range-opening routines" used before. Could you explain what this does or is ????

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Bob Smith"
Wed Jun 6, 2007 8:20 am (PST)

On Jun 6, 2007, at 2:32 AM, Rick Gordon wrote:

I might guess that part of Dan's complaint is that part of the
resulting effect of wide expansion the luminosity range (as might
be required in a scene in low light) -- particularly when there are
big positive moves made with Contrast, Fill Light, or Curves -- is
an increase in saturation, which must then be countered with
various other moves to reduce the saturation back to a desirable
range. It may be all doable, but sometimes the pure luminosity
moves of an L curve or a luminosity blending mode, might
realistically be wished for.

I agree. I just drive these things. I have little idea of what makes them tick. I'll gladly accept that I'm wrong. But it seems to me that when you do a major tonal shift on the raw data before color information is applied that's more akin to a composite RGB curve move than a luminance only adjustment to color image. Possibly that's part of what's happening in ACR. If so, I think that's the source of Dan's comments. When you manipulate the pixel data before the demosaic process takes place, each pixel is treated equally without regard to what color it represents. What might wind up as a fully saturated blue pixel will get just as much correction as a pixel of equal value that winds up being neutral because of the values of its adjacent pixels. A luminance curve would not apply the same move to both of those pixels.

May someone way more familiar with raw processing programming (Iliah?) would care to comment?

Bob Smith

Accurate Image • Bob Smith Photographer • Waco Texas USA
http://www.accurateimage.org
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Stephen Marsh"
Wed Jun 6, 2007 8:23 am (PST)

Rick Gordon wrote:

We probably should wait for Dan to weigh in on what he meant as
"range opening", but my guess is that he is referring to the idea of
expansion (or in the case of Camera Raw, also possibly contraction) of
the luminosity range to cover (in typical images, but not all images)
a shadow point just distinguishable from pure black and a highlight
point just distinguishable from pure white, plus the possibility of
very limited points of pure black or pure white.

Agreed on Dan's reply, and from my scanning experience this was my take on range.

An earlier "hint" from Dan:

http: //tech.groups.yahoo.com/group/colortheory/message/17586

Regards,

Stephen Marsh.
___________________________________________________________________________

Definition of "Range"
Posted by: Dan Margulis
Thu Jun 7, 2007 5:05 am (PST)

I am trying to catch up with list correspondence in more or less the order it came in, but in view of the number of inquiries about the meaning of "range", I will visit that first.

"Range" has been a standard term in the prepress industry for at least a quarter of a century. It can be used to refer either to a picture as a whole or to certain important object(s) contained therein. It denotes how different the lightest and darkest significant areas are.

In preparing an image for reproduction, as noted previously by David Marley in an excellent post, we almost always want to establish endpoints that are as light and dark as the printing process can accommodate with detail. This was historically the job of the scanner operator and/or the dot etcher. This move was described in many different ways, such as setting highlight and shadow, establishing light and dark points, setting range, opening range, establishing range. Photoshop's Auto Levels, Auto Contrast, and Auto Color commands are all range-setters.

In the first edition of Professional Photoshop (1994) I wrote, "The by-the-numbers rules can be stated in a single sentence. Use the full range of available tones every time, and don't give viewers any colors that they will know better than to believe."

I don't have an electronic copy of the 1994 text with me, but in the current edition (PP5E, 2006), the corresponding by-the-numbers chapter (Chapter 3) uses the word "range" 14 times, in contexts like this: "Figure 3.9. In principle, overly dark images can be corrected without human intervention, by adjusting the range automatically.")

The term is used above to refer to the image as a whole, but sometimes it's used to talk about components. If we wish to call more attention to a specific object in the image, the normal way of doing so is by extending its range through curves or channel blending. As stated in PP5E's first chapter on the fundamentals of color correction (Chapter 2 in the book) : "Some images are busy, but many more are not. Product shots, fashion shots, images of animals, food shots—all generally have only one or two color ranges that are important. The rest is just background."

And, " if the advertisement is to be successful, we need to focus the viewers' eyes on the products, just as we needed to focus on [a particular object in the previous exercise]. The method is the same: we extend their range with curves."

All editions of Professional Photoshop have a chapter about range control entitled "The Steeper the Curve, the More the Contrast". All editions label this, "the principal secret to professional color reproduction." The statement remains as true today as it was in the 1980s: without proper range control, consistent professional quality is not possible. What this thread has suggested is that Camera Raw does not offer sufficient control in this critical area.

Again, I have no electronic copy of PP1E at hand, but in the aforementioned Chapter 2 of PP5E, the word "range" appears 27 times.

Dan Margulis
___________________________________________________________________________

Re: Definition of "Range"
Posted by: Mark Segal
Thu Jun 7, 2007 7:18 am (PST)

Dan,

Thanks for responding on the definition of range-opening. Now, you say below the "thread" (who exactly?) suggested that Camera Raw does not offer sufficient control in this critical area. I'm in ACR 4.0 right now, and I count six controls in the Basic tab, seven in the parametric curve dialogue and the full adjustability of the whole tone curve in the Point Curve dialogue - there being however no blending modes given that it is not a rendered image. Having used this tool-set on over a thousand images over the past several months I have yet to encounter situations where "control" has been "insufficient" (however defined), taking into account that a raw converter is not meant to replicate everything one can do in Photoshop - what I'm finding in fact is that the amount of control one now has with this application slices through challenging luminosity issues in ways that were much more difficult in Photoshop proper. About the only thing I need to do with Curves and HSL in Photoshop, after careful use of ACR 4.0, is activate the soft-proof and make final tweaks according to the adjustments so indicated. I realise that your comments are addressed mainly to an outdated version of ACR, but even in CS2 luminosity control was quite substantial. One needs to be careful about defining what one means by "sufficient" - it must have a context and some relativity, remembering that no single application is meant to be the unique answer for every conceivable problem. What's important to me as a user with high standards of (professionally evaluated) image quality is just how much of an efficient and effective workflow I'm achieving with this application. It should not be downgraded with vague generalizations such as "...does not offer sufficient control".

Mark Segal
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: Dan Margulis
Thu Jun 7, 2007 11:00 am (PST)

Matthew writes,

Dan, first of all, great post. I've read the LAB book (although most of this discussion
wouldn't have been appropriate there) but haven't finished all of the latest Professional
Photoshop (I never read any
of the previous editions) so I've seen some of these things scattered
throughout the text, but never put together in one place. Did I miss this?

The topic of what people see and want to see is covered in Chapter 1. As an entire book could be written about that subject, chances are that some of the things we've discussed on this list aren't available there.

Can I ask a chicken-or-the-egg type question? This sort of thing has come up in other
contexts, so I'm curious. Did this research get done
and then Kodak decided to formulate their film for what we now (or
some of us still might call, if we remember) call the "Kodachrome"
look? Or did Kodak decide to formulate their film this way and then
pay for research to back it up?

I don't know what went on in the innards of the Rochester operation, so your guess is as good as mine.

I first heard some discussion about this back in the late-90's when I
bought my first digital camera (a Nikon Coolpix 950). It basically
came down to the Nikon versus the price-equivalent Kodak model at the
time. Of course, there was no RAW mode available on either (Although
the Nikon could create TIFF files, remember when that was an
important feature?) so it was image processing versus image
processing. There were lots of websites (still are) that uploaded
multiple samples of images from both cameras that you can download
and compare and the Kodak was always significantly more red-yellow
than the Nikon, which tended toward blue-green.
The explanation was (and whether this ever came out of a Kodak
engineer's mouth, I don't know) was that Kodak wanted to evoke a
"Kodachrome" look out with the digital cameras, hence this decision.

Again, I don't know, but I know that vendors often make unwarranted assumptions about how much their customers wish to stick to tradition. The most notorious example was the advent of computerized typesetting in the early 1960s. Hot-metal keyboards were deliberately configured in the most inefficient possible way (vowels in the far corners of the keyboard) specifically to slow down the operator, who otherwise might be able to type faster than the machine could process. Much of the first generation of cold-type keyboards--for which there was no technical reason to slow the operator down--used this insane keyboard rather than the QWERTY model, on the theory that probably this is what the customers wanted.

As for why Nikon went for a cooler look, this was attributed to Nikon
being a Japanese company. Someone claimed that when Fuji first
started out, they did the same sort of research and came to the
conclusion that Japanese (and someone else claimed Europeans)
preferred a cooler (blue-green) image than North Americans. Some
thought that Japanese people are more sensitive to greens and blues.

The reverse. They thought they were more sensitive to *reds.*

The more common assertion is that those of Spanish, Italian, or Portuguese ancestry tend to be less sensitive to reds, than would be, say, a group of Swedes. The theory is that, with the sun being so strong in Southern Europe, evolution would have favored those with more visual tolerance of it. Consequently, it has been argued, advertising aimed at Southern European audiences should be warmer than if it is aimed at the Northern climes.

I don't know where you've taught all your classes, but would you say
that there's any truth to any of that or are you even in a position
to make a claim either way outside of North America?

Yes, because I gave the Munsell color test to so many people. I can't say anything about the Southern European theory because I didn't always know peoples ancestry. But I am able to identify when a person is primarily of Asian ancestry. I found that when people had near-perfect color vision (only one error on the Munsell test) persons of Asian ancestry would usually make it in the blues. Everyone else would make it in the greens. The Asians I tested were often Chinese rather than Japanese, but still, I saw a difference.

Dan Margulis
___________________________________________________________________________
 
Re: What makes for 100% approval?
Posted by: "Hoffner, Randall N"
Thu Jun 7, 2007 12:43 pm (PST)

I have no idea whether it is because of genetics, cultural preference, the nature of local sunlight, or any other reasons, but I have certainly noticed that certain colors are frequently used or preferred in certain parts of the world, and these can be colors that are seldom seen elsewhere. This is readily seen in Asia. A couple of examples are that most of the buses in Seoul, at least in 1988, were a kind of Pepto-Bismol color that was frequently seen there, but unfamiliar in the parts of the US that I frequent, or anywhere else I am familiar with. Another example is in Japan, where many public telephones are a very bright, saturated lime-green color that is also not like anything I have seen anywhere else.

Randy Hoffner
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: "Hoffner, Randall N"
Thu Jun 7, 2007 12:44 pm (PST)

To bring this color preference thing a little closer to home, although I haven't heard this for several years, it used to be said frequently that television shows produced in the East (read that New York) had a different color balance from those produced in the West (read that LA). One of them made warmer video, the other cooler video. I don't remember which coast was which. I never heard that film was color balanced for region, though.

Randy Hoffner
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Thu Jun 7, 2007 5:01 pm (PST)

On 6/6/07 8:50 AM, " Rick Gordon" wrote:

Though I am an avid Adobe Camera Raw user and supporter, I might guess that
part of Dan's complaint is that part of the resulting effect of wide expansion
the luminosity range (as might be required in a scene in low light) --
particularly when there are big positive moves made with Contrast, Fill Light,
or Curves -- is an increase in saturation, which must then be countered with
various other moves to reduce the saturation back to a desirable range. It may
be all doable, but sometimes the pure luminosity moves of an L curve or a
luminosity blending mode, might realistically be wished for.

While we wait for Dan to reply...I thought I would do some experimenting on my own to see what I could see. The tests (and I've got the files-and you can replicate the tests yourself with a raw image of a CC).

I took a raw shot of a Color Checker and processed it from Camera Raw 4.1 in Pro Photo RGB 16 bit. The only image "corrections" were to set the white balance (WB click on the 2nd to brightest grey swatch).

I processed it two ways;
1) with a "linear" curve and
2) with a "strong" curve from the points curve adjustment (the "strong" from the drop down menu).

I brought them into Photoshop with my working space set to Pro Photo RGB.

I duplicated both images and converted them from RGB to HSB using the optional 'HSBHSL.plugin' found in the Photoshop Goodies folder of the optional plug-ins.

I examined the "green" channel which represents the saturation...the two were close, but a "bit" different.

I then took the original "linear" contrast RGB file and applied a composite RGB curve to visually match the results of the second processed "strong" curve image from Camera Raw 4.1. I got pretty close to the same contrast but it was clear that the saturation did jump up. I then took the resulting file and converted it to HSB and looked at the green channel (saturation) again and sure enough, there was a noticeable difference in the saturation.

For yucks, I took the original "linear" curve processed RGB file and converted to Lab and then put a luminance based contrast curve to visually match the "strong" curve processed from Camera Raw 4.1.

The result of the Lab contrast curve was different-a bit-from the one with the strong contrast from Camera Raw. It was a "bit" less saturation boosted but not by too much. But both were _CLEARLY_ less saturation boosted than the linear processed with an RGB adjusted increased contrast.

I took the Lab adjusted image and converted to HSB (note, you will have to round trip from RGB>Lab>RGB to do the HSB conversion-no choice there). I looked at the saturation channel (green remember) and compared it to both the Camera Raw 4.1 increased contrast and the Photoshop RGB curve contrast images...

The Lab luminance adjustment and the Camera Raw "strong" contrast adjustment saturation looked relatively close. A lot closer to each other than the Photoshop RGB contrast adjustment.

What's my conclusion? That Camera Raw's contrast adjustment, while not _EXACTLY_ like a luminance curve adjustment in a Lab image, looks _MORE_ like a Lab luminance adjustment than an RGB "Master" contrast adjustment.

And that Camera Raw's curves are pretty close to luminance curve adjustment and better than a master channel RGB contrast adjustment.

You can test this yourself...(no need for Excel, but ya need a raw image of a ColorChecker).

Regards,
Jeff Schewe
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: Jeff Schewe
Thu Jun 7, 2007 5:02 pm (PST)

 "Hoffner, Randall N" wrote:

To bring this color preference thing a little closer to home, although I
haven't heard this for several years, it used to be said frequently that
television shows produced in the East (read that New York) had a
different color balance from those produced in the West (read that LA).
One of them made warmer video, the other cooler video. I don't remember
which coast was which. I never heard that film was color balanced for
region, though.

Pretty easy to see the bias...look at the CSI series. CSI LV is warm but tends to be more golden than say CSI Miami which is both warm and punchy colors...then look at CSI NY. It's usually got a strong cool bias almosy cold/blue.

And believe me, these are very intentional production design decisions...nothing happens by accident on TV series or feature film production. Lice TV, yes...stuff "happens", but not when there's a whole production design team of art directors and designers working on production.

Jeff Schewe
___________________________________________________________________________

Re: What makes for 100% approval?
Posted by: "Hoffner, Randall N"
Fri Jun 8, 2007 10:31 am (PST)

You got that right. People who don't know how this works tend to blame the post house when something looks "different" or "not right", "too dark", etc., but it doesn't get out of the post house without the producers' approval.

Randy Hoffner
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Fri Jun 8, 2007 10:53 am (PST)

Jeff Schewe writes,

While we wait for Dan to reply...I thought I would do some experimenting on
my own to see what I could see. The tests (and I've got the files-and you
can replicate the tests yourself with a raw image of a CC).

I took a raw shot of a Color Checker and processed it from Camera Raw 4.1 in
Pro Photo RGB 16 bit. The only image "corrections" were to set the white
balance (WB click on the 2nd to brightest grey swatch).

This is not a very good test because the swatches in a Color Checker contain no detail, so the damage cannot be readily seen. Also, the colors themselves are not particularly appropriate. I therefore didn't repeat the test but the conclusions seem logical.

I agree that it would be difficult to determine from it whether the curve being applied is a master curve to overlaid RGB channels, or a luminance curve. Fortunately, there are images that will demonstrate it conclusively one way or another.

I am glad that Jeff has found out what "master curve" means. In investigating, he has undoubtedly learned that this is considered an inferior way to correct images even granted that other ways are provided, and that it's completely unacceptable for it to be the only option for a professional-level program.

That is, IF an RGB master-curve structure is really being used in the ACR Curve command. Jeff's bosses apparently inform him that it is not so. Instead, he is informed, the Curve is based on luminosity, rather than RGB-like channels.

As Jeff noted, testing a Color Checker doesn't give a clear answer as to whether they are right. What we need is an image that *will*--that is, an image where the difference between a luminosity correction and an RGB master curve is so decisive and so obvious that nobody can deny which is which. If anyone else wishes to do the test themselves, here would be the steps.

1) Must start with a raw or uncorrected image featuring a large, red, detailed object or objects. The red should be deep, but not totally brilliant. A red rose might make a good start. Do not use something orangy-red like a fire engine, or something that is red but has no detail, like a stop sign.

2) In ACR, set range, if necessary, using the Exposure, Brightness, and Shadows sliders, to satisfaction. Do not adjust saturation or anything else because it may make the test harder to evaluate.

3) Save a copy of this precorrected version for future comparison.

4) Returning to the ACR version, apply a strong S-shaped curve using the Curve command, and open the image.

5) Compare the version prepared prior to the curve to the one afterwards.
A) If the curve command applies to luminosity, as apparently claimed by Adobe, the second version will appear to have gained contrast in the red object(s). Certain parts of the object(s) will have become a deeper, but not a brighter red.
B) If the curve is in the nature of a master curve applied to RGB channels, there will be an apparent loss of contrast in the lighter areas and, depending upon the image, possibly in the darker ones as well. Aslo, the lighter areas will appear redder.

If there is any difficulty interpreting this, I have posted a JPEG image suitable for testing at

http://colortheory.smugmug.com/gallery/2971728

I have also posted one version that was corrected in Photoshop with an RGB master curve, and another by applying the same curve to the L of an LAB version of the file. Granted that the math is different in Camera Raw and Photoshop, these two variant images are as different as chalk and cheese, so there should be no difficulty whatever deciding which one your own version, when you test in Camera Raw, resembles.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Fri Jun 8, 2007 5:52 pm (PST)

---Dan Margulis wrote:
http://colortheory.smugmug.com/gallery/2971728
 
 I have also posted one version that was corrected in Photoshop with an RGB master
curve, and another by applying the same curve to the L of an LAB version of the file.
Granted that the math is different in Camera Raw and Photoshop, these two variant images
are as different as chalk and cheese, so there should be no difficulty whatever deciding
which one your own version, when you test in Camera Raw, resembles.

I'm working on the rest of the response to this post, but I can tell you that doing it Dan's way, bringing the untagged and presumed sRGB jpg image into Camera Raw 4.1, it sure looks ALOT more like the Lab corrected image than the Photoshop composite RGB curve adjusted image...unless I'm doing something wrong.

I set my PS prefs to "Prefer Camera Raw for JPGs" and popped it into Camera Raw 4. My Camera Raw 4.1 workspace is set to ProPhoto RGB in 16 bit. I took the original "flatter" first image & applied a "Strong Contrast" curve from the drop down menu in the Tone Curve panel (I preferred to choose a standard drop down default so as to allow anybody else to verify the curve since you can't save out curves from CR).

By default, jpgs in Camera Raw 4.x come in with a linear (flat) curve...it looks pretty much like what the image looks like in a browser (and in Photoshop when the image is assigned sRGB). I applied the "Strong Contrast" curve and compared it to the Lab adjusted image...looks pretty close-not EXACT mind you but pretty close. But NOTHING like the one that Dan says was Photoshop adjusted...I COULD get the original jpgs to look like that in Camera Raw, but it required both a strong curve AND a saturation setting boost of +18.

Note, I did not change any other Camera Raw settings in this quick test...just the Tone Curves and the Saturation setting (to match Dan's Photoshop result).

So, am I wrong? Doesn't this visually prove that Camera Raw is _NOT_ doing the same thing as Photoshop? Dan writes, on the initial page: "This gallery is an aid to those wishing to test whether the Curve function in Camera Raw uses as its model a master curve based on RGB-type channels, or whether it is based purely on the luminosity of the image".

Well, I can't get the Camera Raw adjustments to match the Photoshop adjustment without the CR sat boost. So, I'm pretty sure this proves that Camera Raw is _NOT_ using "as its model a master curve based on RGB-type channels". Unless I screwed up the test...

So, Dan, what did I do wrong? Why doesn't the Camera Raw curves adjustment look like Photoshop's composite RGB curves adjustment?

Regards,
Jeff Schewe
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Rick Gordon"
Fri Jun 8, 2007 6:50 pm (PST)

I can't shift gears to test this right now, but is there any possibility the "Blend RGB Colors Using Gamma ___" in the advanced color settings (which I usually have enabled and set to 1.00) could make a difference in the results of an RGB master curve?

Rick Gordon

___________________________________________________

RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________________

WWW: http://www.shelterpub.com
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Fri Jun 8, 2007 6:50 pm (PST)

Jeff Schewe writes,

I'm working on the rest of the response to this post, but I can tell you that doing it Dan's
way, bringing the untagged and presumed sRGB jpg image into Camera Raw 4.1, it sure
looks ALOT more like the Lab corrected image than the Photoshop composite RGB curve
adjusted image...unless I'm doing something wrong.

I set my PS prefs to "Prefer Camera Raw for JPGs" and popped it into Camera Raw 4. My
Camera Raw 4.1 workspace is set to ProPhoto RGB in 16 bit.

The file is in fact tagged, but I have spoken to Jeff offline and he has stated that he has redone the test with the same result.

I took the original "flatter"
first image & applied a "Strong Contrast" curve from the drop down menu in the Tone
Curve panel (I preferred to choose a standard drop down default so as to allow anybody
else to verify the curve since you can't save out curves from CR).

No, but you can tell people what your I/O points are. I specified a strong S curve, which the "Strong Contrast" curve is not.

By default, jpgs in Camera Raw 4.x come in with a linear (flat) curve...it looks pretty much
like what the image looks like in a browser (and in Photoshop when the image is assigned
sRGB). I applied the "Strong Contrast" curve and compared it to the Lab adjusted
image...looks pretty close-not EXACT mind you but pretty close. But NOTHING like the one
that Dan says was Photoshop adjusted...I COULD get the original jpgs to look like that in
Camera Raw, but it required both a strong curve AND a saturation setting boost of +18.

Note, I did not change any other Camera Raw settings in this quick test...just the Tone
Curves and the Saturation setting (to match Dan's Photoshop result).

The test specifically stated that Saturation should not be touched. It was specifically stated that there would not be a match to either of my Photoshop versions.

So, Dan, what did I do wrong? Why doesn't the Camera Raw curves adjustment look like
Photoshop's composite RGB curves adjustment?

Conceivably because there is an operator error, but rather more likely because we tend to see what we want to see.

I have made a new gallery at
http://colortheory.smugmug.com/gallery/2973533
consisting of two ACR-processed files, as follows.

The files were prepared on a brand-new install of ACR 4.1. No image had ever been processed with it before. The file was opened as sRGB and saved out, with the only change being a Custom curve with two internal I/O points at 64=33 and 192=224. This was saved as version 1. It was then reopened and version 2 made with the points made less radical at 64=45 and 195=215.

The results I predicted were that if a master-curve structure was in place, there would be an obvious increase in saturation in the lighter areas and an obvious loss of detail in the lightest areas as well, factors that would be absent in a luminosity-based version.

I am. as you are, a biased observer of this test. Therefore, I had a jury look at it--the class I'm currently teaching, seven serious people, none of whom knew what is going on here or why I was asking the question. I showed them Version 2 (as being closer to my Photoshop versions) layered over the luminosity version and the master-curve version. Without permitting them to express any view to one another, I repeatedly toggled back and forth between the two variants, asking, first, which one was closer to the ACR version 2, and whether they considered it obvious. I permitted them to ask to see further comparisons of either version against the ACR version, but not to explain why they were asking for it to avoid biasing other persons. After hearing the vote, I asked further questions of the group and again asked whether the answers were obvious.

The vote:

Which version is closer overall to the ACR version?
5 persons: it is obvious that the master curve version is closer.
2 persons: we feel that the luminosity curve version is closer, but it is a close call.

I then asked the two persons choosing the luminosity curve version why they felt the way they did. They answered that the master curve version is extremely flat. and that from the artistic POV there is an impression of snappiness in ACR version two that is matched by the luminosity version better than by the master curve version, even though in certain respects the match is not close at all.

I then asked the questions with respect to my predicted result that there would be an clear increase in saturation compared to the luminosity version, and that there would be an obvious loss of contrast in the lighter areas. The vote is unanimous that these factors are there and that it is obvious that in these respects the ACR Version 2 is much closer to the master curve version than it is to the luminosity version.

I did not show them Version 1, which uses the stronger curve, because it is screamingly obvious that it is even more saturated than Version 2 is (luminosity moves shouldn't change saturation at all) and that the loss of detail in lighter areas is even more pronounced.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Mike Russell"
Fri Jun 8, 2007 10:01 pm (PST)

From: "Jeff Schewe"

I'm working on the rest of the response to this post, but I can tell you that doing it Dan's
way, bringing the untagged and presumed sRGB jpg image into Camera Raw 4.0, it sure
looks ALOT more like the Lab corrected image than the Photoshop composite
RGB curve adjusted image...unless I'm doing something wrong.

I did the experiment as well, using sRGB and a somewhat stronger S curve and I agree with Jeff's result. The result is visually more similar to the Lab Lightness manipulation than to the RGB composite one.

I added the Lab image as a layer to the RGB result, and set the opacity to 50 percent. The result is visually similar to the result from ACR 4.0. From a saturation standpoint, and judging from Dan's cherry image only, ACR is using something about halfway between the RGB composite curve and the Lightness curve of Lab.

I dug up the following comment from John Knoll, made on Oct 11, 2005 in the camera raw group at adobeforums.com:

"While developing Camera Raw, I experimented with a pure luminance curve (as Simon [a previous poster] suggests). However, based on my testing results, I rejected this algorithm since it produced results that were most often visually worse looking that the tone curve algorithm actually used by Camera Raw (which is a special hue-preserving curve, NOT three indepent curves as Simon incorrectly assumed). The saturation effects that Simon considers a defect is actually something that most users actually want. "

So, ACR's curve may be a hybrid that links saturation and contrast together while retaining hue. My preliminary findings show that not all hues are preserved in all situations: ACR 4.0 preserves some hues, and changes others, particularly greens.

Mike Russell - www.curvemeister.com
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Iliah Borg"
Fri Jun 8, 2007 10:01 pm (PST)

unless I'm doing something wrong.

Have you considered converting to linear Prophoto colour space in Photoshop prior running composite curve on the image? Apples to apples?

--
Best regards,
Iliah Borg
___________________________________________________________________________

Blend Using Gamma 1 & Curves (Originally: Color theory and practice)
Posted by: "Stephen Marsh"
Fri Jun 8, 2007 10:02 pm (PST)

Rick Gordon wrote:

I can't shift gears to test this right now, but is there any
possibility the "Blend RGB Colors Using Gamma ___" in the advanced
color settings (which I usually have enabled and set to 1.00) could
make a difference in the results of an RGB master curve?

Going from memory, no. Blending occurs with layer opacity or blending opacity sliders and with pixel averaging such as blurs. So adjusting the opacity of a curve adjustment layer may call this into play if the curve was applied to pixel data and the pixel data blended into underlying pixel data in a layer. But not directly to the curve/image itself. I do not recall that the option affects say a curve adjustment layer (even if at reduced opacity), the values were the same toggling this gamma blend option on and off.

For some images, edits and workflow, there would be no difference between using this setting or not, as blending does not occur. Where blending does occur, it would be image and viewer dependent on which result was preferred.

Do you really mean enabled as a general working default Rick? The app defaults with this off/disabled (probably for legacy issues and or meeting most users expectations of editing response). So you made a conscious decision to work with this on? What lead to this choice? The results? Do you layer stack and blend a lot, or do large blurs or 'paint' in Photoshop? When comparing your workflow with your common images with this option both on and off, you prefer it on?

The list once had a good thread about this advanced colour settings Blend RGB function, with regards to an averaging colour test suggested by Timo at AIM using a two colour CG pattern. I found that various gamma RGB, dot gain CMYK and LAB had different advanced blending results (I think I preferred LAB and CMYK over RGB, but I can't recall the situation in more detail). Yes, 2.2 gamma did turn hues neutral with a large average, but I did not agree that gamma 1 was better than another RGB gamma choice, LAB or CMYK - just different. Many images and blending edits would need to be tested and compared before I could say which was a better default.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Sat Jun 9, 2007 5:44 am (PST)

Dan Margulis wrote:

I have made a new gallery at http: //colortheory.smugmug.com/gallery/2973533
consisting of two ACR-processed files, as follows.

The files were prepared on a brand-new install of ACR 4.1. No image had ever been
processed with it before. The file was opened as sRGB and saved out, with the only change
being a Custom curve with two internal I/O points at 64=33 and 192=224. This was saved
as version 1. It was then reopened and version 2 made with the points made less radical at
64=45 and 195=215.

Uh huh...so, I'm confused (I often am you know, which is healthy when it comes to learning)...You state, on the SmugMug site: "These files were processed on a fresh install of ACR 4.1. The only change upon opening was that to each version a Custom S curve was applied, having only two adjustment points. In version 1, the adjustment Input/Output points are 64=33, 192=224. In version 2, the points are 65=45. 195=215."

OK, but when I download both files and look at them in Bridge, this is what I see in the metadata...

File name: 160855923_Cherries_dm_acr_version2.jpg
Date Created: today, 8:18:17PM
Color Profile: sRGB IEC61966-2.1
Application: Adobe Photoshop CS3 Macintosh

File name: 160855900_Cherries_dm_acr_version1.jpg
Date Created: today, 7:35:33PM
Color Profile: sRGB IEC61966-2.1
Application: Adobe Photoshop CS2 Macintosh

So, what did you do to the file while it was in Photoshop CS2? Pretty sure Camera Raw 4.1 only works in CS3. So, did you open it in CS3, save and then reopen in CS2?

It might be useful, just so your exact steps can be duplicated that you turn on history logging so we can follw exactly what you do to the file.

It should also be noted that what we're doing...using Camera Raw on JPGs certainly wasn't what you had done previously when you assembeled your spreadsheet data, correct? In the grand scheme of things, using Camera Raw on jpgs, is, well kinda an odd way of using Camera Raw...but, whatever. However, I did process your image with CR's curve using 64=45 and 195=215. Then processed the original jpg with linear (no curve) and applied the same curve points in Photoshop. The Camera Raw curves does not match the appearance of the Photoshop curve version.

I also took the original jpg, converted to Lab and tried to match the curve (course I can't use the exact same input/output numbers since Lab ain't 0-255) and looked at all three images:

Lab curve: (input: 24, output: 16 & input: 75, output: 87) color looks flat and a bit unnatural but the contrast increase had zero effect on saturation...

Camera Raw curve 64=45 and 195=215 increased the contrast and has an effect, mild, on saturation...

Photoshop CS3 curve (I mention that fact it's CS3 since the CS3 curve tool _IS_ different than previous versions), the contrast increase comes with considerable increase in saturation.

Which is "better"? Well, that is certainly open for debate. But what is NOT open for debate is the statement that Camera Raw uses a "Master Curve" approach that is the same as Photoshop (at least in CS3).

Regards,
Jeff Schewe
___________________________________________________________________________

Re: Color theory and practice
Posted by: "George Machen"
Sat Jun 9, 2007 11:44 am (PST)

I don't know about Camera Raw 4.1, but for the version in Photoshop CS2, as I stated here last February, ...

<http: //tech.groups.yahoo.com/group/colortheory/message/16738>

"... while Bruce Fraser's Camera Raw book does contain one sentence stating that Camera Raw uses an algorithm to preserve color shifts during such master adjustments (like a Photoshop Color layer mode or Fade command, I guess), that still leaves the obliteration of tonal detail & contrast from an overall move (i.e., there isn't an algorithm providing a constraint analogous to a Photoshop Luminosity layer mode or Fade command ... and even if it did, what further adjustment then could be possible?)."

(That one sentence in the late Bruce Fraser's book is the last sentence of p.35.)

Bruce Fraser grasped the significance of the folly of master-channel movements!

Workarounds exist for the damage done by composite master channel adjustments, such as doing so in Luminance, Color, Hue or Saturation mode(s), but some damage remains. If one invokes all the blending modes, it's a trivial result: nothing at all!

The upshot is that there's no way around making individual-channel movements with impunity. (Except for neutral-color significant subjects, which is a moot point.)

- George Machen
___________________________________________________________________________

Re: Color theory and practice
Posted by: "David Marley"
Sat Jun 9, 2007 11:45 am (PST)

Thank you, Mike, for your research on the history of this feature.

I compared the results of Dan's test using his preferred method of layering different versions into a single file. To isolate the saturation changes, I converted the file to LAB mode (without flattening], and placed a curve adjustment layer on top. I altered the L-channel by setting both the white and black points to 50%, thus elimintaing all luminosity detail. This may help make the extent of the color changes more visually obvious.

Also, thank you, Dan, for your continually thorough and unbiased analysis of Photoshop's features and behaviors.

Dave Marley
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Sat Jun 9, 2007 2:31 pm (PST)

George,

Please quote the full sentence from Bruce Fraser exactly as he wrote it and provide a bit of the context within which he wrote it for the benefit of those such as myself who don't have the second edition of his book. For starters, I would like to better understand which aspects of the stated position are his and which your's, and I would like to know what valid evidence is provided backing-up both.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Sat Jun 9, 2007 5:20 pm (PST)

George,

Further to my post below, I have now obtained the sentence you mention from Bruce Fraser's Camera Raw book (second edition):

Here it is:

"The Camera Raw controls use an algorithm that preserves the original hue, whereas hard curve adjustments to the composite RGB curve in Photoshop can cause slight hue shifts."

That is ALL.

I have scoured the contents of Chapter 6 "Real World Photoshop CS2" written by Bruce Fraser and David Blatner, and there is NOTHING in that whole chapter which raises the spectre of the dire results you ascribe to adjusting images with the RGB composite curve rather than doing so with each individual curve. Surveying other well known, published Photoshop instructors, Michael Kieran is also an advocate of channel by channel adjustments ("Photoshop Color Correction", Peachpit 2003). Other prominent authors such as Ben Wilmore and Martin Evening recommend making general tonal adjustments using the RGB composite Curve and using the individual channel curves for correcting color casts.

I think it makes sense to take a nuanced approach to this issue, along the lines that the preferred editing technique (RGB composite versus individual channels) depends on the image and the objectives for that image. Based on much experience using these tools, I don't think it makes sense to rule out either if it serves its purpose properly, and RGB composite curves often do just that very well, while operating individual channel curves may also be most appropriate in other conditions, as Dan and Michael K have demonstrated.

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: "George Machen"
Sat Jun 9, 2007 7:20 pm (PST)

MARK SEGALwrote:

Please quote the full sentence from Bruce Fraser
exactly as he wrote it and provide a bit of the
context within which he wrote it for the benefit
of those such as myself who don't have the second
edition of his book.

Here is the one sentence - the last sentence in the first paragraph - within its entire context, viz., quoting the entire section (pp. 33-36):

"Brightness and Contrast
"The Brightness and contrast controls let you tweak the conversion of the intermediate tones from the linear capture to the gamma-corrected output space. They work completely differently from the similarly named Photoshop Contrast and Brightness controls. Instead, they behave similarly to Photoshop's Levels and Curves, respectively. (Brightness is a midtone adjustment, Contrast is an S-curve) but with one important difference. The Camera Raw controls use an algorithm that preserves the original hue, whereas hard curve adjustments to the composite RGB curve in Photoshop can cause slight hue shifts. "If you make little or no adjustment with the Exposure slider, it's mildly advantageous to use Camera Raw's Brightness and Contrast sliders rather than using Photoshop's tools. But the bigger your Exposure adjustments, the more essential making matching Brightness and Contrast moves becomes (see Figure 2-B , earlier in this chapter."

For starters, I would like to better understand
which aspects of the stated position are his and
which your's, and I would like to know what valid
evidence is provided backing-up both.

A lot of people say they've never seen any problem applying composite channel adjustments. Well, you've seen the equivalent reduced problem in a grayscale if you've ever applied the same curve movement to separate images of three cats - white, gray and black. Images of the deleterious consequences of precisely that dubious practice appear on p. 32 of Dan's PP5E.

And just yesterday Dan posted on SmugMug comparative images of red fruit vividly illustrating the same problem when expanded to color images:
<http: //colortheory.smugmug.com/gallery/2971728#160746476>

The important thing is that in a color image, at least two of the three cats separately occupy the individual channels. In the case of Dan's berries, the white cat is in the red channel, the gray cat is in the green channel, and the black cat is in the blue channel. Apply any curve to the master composite, and you've improved one cat and hosed at least one other.

Or say you apply a midtone S-curve to a wedding shot. Yeah, the image looks better at first glance - more snap - but look what you've done to the shadow detail in the groom's black tux, and to the delicate highlights in the bride's gown, arguably the very objects of crucial significance!

Extended to color objects, sacrificing detail in them in the name of more snap in the overall shot with a master channel move is shortsighted. Gaining detail in colored objects via individual curves can still give the overall image more snap, too, as can be seen in Dan's red berries.

For the life of me, this is not controversial, and has been well-known by prepress scanner operators for decades!

- George Machen
___________________________________________________________________________

There is, IMHO, considerable valuable material in this file.The interlocking threads that have been edited to make it up, however, were noteworthy for  an excessively low signal-to-noise ratio. You have read a small fraction of the offending posts (and responses thereto) above. Many of the others that provoked the acting list moderator to post the following have been deleted. Throughout the rest of this file, the same policy applies: postings devoid of technical interest and/or consisting primarily of personal attacks, and responses to such posts, have been eliminated, as have all but the final moderator statements. If you wish to read the full material, it’s at
http://www.ledet.com/margulis/misc/ACR_Thread_Full.pdf
Warning, however: it’s a 410-page, 1.3 mb text-only document. That's roughly twice the length of what you're reading now. --DM

Professionalism, courtesy on this list
Posted by: "John Ruttenberg"
Sat Jun 9, 2007 6:18 am (PST)

I'm totally bored with the unprofessional, uncourteous, and personal posts on this list. It just adds noise to the list without adding any value. How can anyone compare us in terms of professionalism to dgrin? People there are generally polite, helpful, respectful of the opinions of others, and willing to acknowledge the limits of their own knowledge. By contrast, we should be ashamed of ourselves.

I was so bored with the situation yesterday morning that I rejected several posts, including some I agree with, without consulting other list moderators. Now that I have gotten messages of support from those I've been able to contact, let me state the following:

1. Our rules already state that the moderators reserve the right to reject posts devoid of technical content. IMHO several of the recent posts that were approved should not have been.

2. I WILL reject posts which I find to be personal in nature, unprofessional in tone, or generally not courteous, and I hope that other moderators will do the same.

3. The list exists so we can learn from each other. It behooves every poster, even Dan, to make an effort to understand the point of view presented by others before responding to it. Shouting and repetition do not persuade. The moderators reserve the right to reject posts which show a lack of respect for other people's well considered opinions.

Oh, and I don't encourage onlist discussion of this topic. Contact me by private email if you want to discuss. Enough meta discussion. Let's discuss image correction and enhancement technique and see how it applies in practice. Let's see if we can distinguish ourself with what we can do and how helpful we can be to others trying to improve their own craft.

--John Ruttenberg
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Sat Jun 9, 2007 5:19 pm (PST)

On 6/9/07 7:17 AM, " Dan Margulis" wrote:

I am glad that Jeff has found out what "master curve" means. In investigating,
he has undoubtedly learned that this is considered an inferior way to correct
images even granted that other ways are provided, and that it's completely
unacceptable for it to be the only option for a professional-level program.

To be honest, I've always heard it called a composite curve (working on all three or 4 channels) as apposed to single channel curve-not a "master curve". So, when somebody uses nomenclature that is unfamiliar, I always like to make absolutely sure what somebody is trying to say. Clarity is useful...

That is, IF an RGB master-curve structure is really being used in the ACR
Curve command. Jeff's bosses apparently inform him that it is not so. Instead,
he is informed, the Curve is based on luminosity, rather than RGB-like
channels.

That's an ad hominem argument...two things wrong it; 1) the only "boss" I have is my wife of 33 years-because although I do work WITH Adobe from time to time, I don't work FOR Adobe and 2) my friends at Adobe hadn't "informed" me about anything with regards to Camera Raw's curves...I read it in a book.

From Bruce Fraser's Real World Camera Raw, page 35, Bruce writes under the Tone-Mapping Controls section: "The Camera Raw controls use an algorithm that preserves the original hue, whereas hard curve adjustments to the composite RGB curve in Photoshop can cause slight hue shifts."

So, my "bosses" (friends) hadn't "informed" me about curves before...Bruce did-and he wasn't my boss, he was my friend and partner.

The fact that I believed that paragraph above meant it was a luminance based adjustment in Camera Raw was my own deduction-not entirely accurate I now find. And I _LOVE_ to find out when I'm in error because it means I learned something when I do.

Since my original post back when I did my ColorChecker test I HAVE had some communications with my "bosses" (friends) to find out exactly what Camera Raw's curves were based on (it's always useful-if you want to know the truth-to ask the guy that wrote the code...)

And, as Mike Russell pointed out, Thomas didn't use a pure "luminance only curve" but a hue locking, saturation tuned luminance curve. See, even though using JUST a luminance based curve would have been simpler and more efficient code, Thomas went to the trouble of adding a saturation tweak because he said:

"I have tried "pure luminance curves" in experiments. The math is actually simpler and faster than what I'm using actually using, but the results look bad (to be polite) in most cases.

It maps the min and max of the RGB values (in linear ProPhoto RGB) through the tone curve, and the middle value is interpolated to exactly preserve hue. This results in saturation effects that closely match saturation effects you get from real film tone curves (which many years of looking at film photos has taught our eyes to like), while preventing the adverse hue shifts you would otherwise get from the concave and convex parts of the curve."

He also went on to state: "The exact math we use for our tone curve application is actually published in the routine "RefBaselineRGBTone" in the DNG SDK."

So, anybody can download the DNG SDK, open the source code and see EXACTLY what's being done.

So, Dan...I suggest you retreat from the position that Camera Raw uses a "master curve" function, cause it doesn't.

And while you may think you understand my relationship with Adobe...but it's clear, you don't. Yes, I did consult with Mark Hamburg and Thomas Knoll for the purpose of improving the sharpening in Camera Raw...but that was Bruce Fraser's original contract-which sadly, I had to fill in for. I think Camera Raw 4.1's sharpening is pretty way cool-I didn't have anything to do with the curves (although I did defend the parametric curves version that Hamburg developed and fought to keep them in Lightroom-and eventually into Camera Raw).

I also promised Bruce that I would take over his Real World Camera Raw book as well. So, that puts me in the position of being kinda the Camera Raw "Watcher" (making sure stuff people say about Camera Raw is true and accurate-and finding out the truth when they or I are in error).

In that role, I welcome learning anything and everything I can as it relates to Camera Raw...and if _YOU_ have questions regarding _EXACTLY_ how Camera Raw works, you just let me know...I'll pass the questions along. I do have friends (but no bosses) that are inclined to tell me the truth...and isn't that ultimately what this is all about...finding out the truth?

In this case, the truth is Thomas make an informed decision to do a hue locked, saturation tweaked, luminance curve adjustment because he thought that worked better.

I agree. A pure luminance curve adjustment is TOO "un-photographic" and doesn't look natural but Camera Raw's curves adjustments ARE better than Photoshop's composite curves adjustment (and different)...

Regards,
Jeff Schewe
___________________________________________________________________________

Re: [ Color theory and practice
Posted by: "George Machen"
Sun Jun 10, 2007 5:57 pm (PST)

Jeff Schewe wrote:

So, Dan...I suggest you retreat from the position
that Camera Raw uses a "master curve" function,
cause it doesn't.

Wait a minute, indeed it does, and ironically by your own argument (or more precisely, by Thamas Knoll's), quoted below.

While the described ACR semi-Luminance-mode curve "[prevents] the adverse hue shifts you would otherwise get from the concave and convex parts of the curve," that's not all there is to the problem with master-composite channel moves: it still doesn't prevent the second aspect of loss of detail & contrast in at least fairly brightly-colored objects that can, and *only* can, be avoided via application of individual channel curves. If it's not an individual-channels move, it's still essentially a composite-channels move, no matter what mode-preserving function is applied.

This second aspect was the main point described in my immediately prior post:
<http: //tech.groups.yahoo.com/group/colortheory/message/17979>

By the way, Jeff, it's delightful to have you on board here at ACT. I think your future contributions will be of inestimable value!

- George Machen
___________________________________________________________________________

Re: [ Color theory and practice
Posted by: Jeff Schewe
Sun Jun 10, 2007 8:59 pm (PST)

"George Machen" wrote:

Wait a minute, indeed it does, and ironically by
your own argument (or more precisely, by Thamas
Knoll's), quoted below.

Reread what he said...if locks the hue and tempers (tunes) the saturation...there's no way you can call THAT a "Master Curve" adjustment...the very fact he's locking Hue and modifying the way saturation is applied means it's conceptually and fundamentally different that applying a single composite curve as done in Photoshop's curve...

And not only is it NOT a composite curve nor a straight luminance curve, he's developed an algorithm that is unique to Camera Raw...one that makes raw processing more like film rendering...

Dan said it was a "Master Curve". He said he ran tests, charted the results in a spreadsheet and has proof that it's a Master Curve. Thomas said it isn't...who ya gonna believe? I believe Thomas and if YOU want, go download the DNG SDK, pop the code into Xcode and see EXACTLY what the curves are doing-you can do that. I don't really have a head for math...

Reread my description for more clarification: "the truth is Thomas made an informed decision to do a hue locked, saturation tweaked, luminance curve adjustment because he thought that worked better."

That ain't no Master Curve bud...no way you can lock hue, tune the saturation and also apply a luminance curve in a single Master Curve operation.

And run Dan's own tests on his image (and on his terms) and you'll see that Camera Raw doesn't match Photoshop's curves adjustment...

So, I think that puts it to rest. It ain't a master curve...ok?

Now, could it be better? I don't know. I kinda like it the way it is...I have no idea how many images I've run through Camera Raw since version 1.0, but I suspect it's more than Dan has. I really like Camera Raw's rendering of raw images...I like the cotrols as they are now in 4.x. I particularly like to use the parametric curves. (of course I WOULD still like to get localized controls)

But, if you have a problem image that just can't be fixed in Camera Raw, there's this little application out there called Photoshop-you may have heard about it. You can even take an RGB image and convert to Lab and run a pure luminance curve on the Lightness channel too. I hear there'a guy that wrote a whole book on the subject of Lab...

:~)

Regards,
Jeff Schewe
___________________________________________________________________________

Re: Definition of "Range"
Posted by: Dan Margulis
Sun Jun 10, 2007 9:00 pm (PST)

Mark Segal writes,

Now, you say below the "thread" (who exactly?) suggested that Camera
Raw does not offer sufficient control in this critical area. I'm in ACR 4.0 right
now, and I count six controls in the Basic tab, seven in the parametric curve
dialogue and the full adjustability of the whole tone curve in the Point Curve
dialogue - there being however no blending modes given that it is not a
rendered image.

The point is not how many commands have been added, but how well they can address the specific problems that have repeatedly been pointed out to you in specific images--problems that any professional-level program from around 1985 on does not encounter. As far as I can determine, for all the nice new functionality that has been added in CS3, the fundamental problems in range extension have not been remedied.

One needs to be careful about defining what one means by "sufficient" - it
must have a context and some relativity, remembering that no single
application is meant to be the unique answer for every conceivable problem.
What's important to me as a user with high standards of (professionally
evaluated) image quality is just how much of an efficient and effective
workflow I'm achieving with this application. It should not be downgraded with
vague generalizations such as "...does not offer sufficient control".

The context is for a professional user, specifically one whose livelihood depends on meeting the particular requests of clients. You have been referred to at least five example images where the insufficient control is demonstrated. In each case, the image problem was either verified by several other people or, once, by inaction on the part of the others. I have given you a technical explanation of why it occurs and what the ramifications are. One has to ask, if five image examples are "vague generalizations" that a problem occurs, how many more would be necessary?

Just to remind you of these examples, what the claim is, and how it has been verified:

1) Image of red flower, in PP5E. Claim: that establishing a proper range in ACR could not be implemented without knocking out so much detail in the weak channel that the file could no longer be corrected properly. Correction shown of a version acquired without setting range in ACR, and statement made that the result can't be matched if range is opened in ACR first. Verification: by 10 beta readers, plus by implication by the fact that no reader of the book has offered any other solution, as they commonly do when a mistake is found.

2) Image of indoor shot of couple, in PP5E. Claim: that establishing range in ACR unreasonably whitens a light point that is supposed to be yellow. Ease of correction shown when the file is acquired without setting range. Demonstration that the damage cannot be contained by use of the ACR White Balance routine. Verification: as above.

3) JPEG image of architecture, by Boris Feldblyum. Claim: that the JPEG algorithm, which is very similar to the one used in ACR, created a correction nightmare by oversaturating conflicting colors. Suggestion that if the file had been shot raw, it would have been easy to correct as long as the ACR range-opening routine wasn't used. Use of ACR saturation correction not possible due to the casts conflicting with real colors. Verification: that several members of this list attempted to correct the image, and, unless I missed one or two, none were professionally acceptable. After this, since no raw file was available, I posted the numerical results of setting range in ACR given a file with similar colors, clearly showing the oversaturation being complained of in Boris's image.

4) John Ruttenberg's ballerina picture. Claim: As soon as the image was posted I commented that one would not be able to extend the range in ACR without so much damage that a channel blend would likely be needed to correct the color. Verification: that several members of this list *did* open the range and the only ones who were successful *did* have to resort to channel blending, whereas the ones who acquired a flatter image did not.

5) Same picture. Members of this list posted three or four examples that the photographer (and I, and everyone else who commented on them) thought were good renditions. Claim: that none of these straightforward global moves could have been achieved in ACR by any combination of commands. Verification: that the Camera Raw cheering section, which ordinarily posts around 25 rebuttals to anything I write, was nowhere to be found, whereas a demonstration that I was wrong would have had a telling impact. You yourself were referred to this image and declined to get involved.

If there is anything vague about any of these examples, please let me know, or let me know how many further examples of damage would satisfy you. Since you claim not to have difficulties with the module yourself, perhaps you would try correcting the images and seeing if you can get better results. In fact, let me suggest yet another example, which you will hopefully not find vague:

6) The gallery that I created at
 http://colortheory.smugmug.com/gallery/2971728
should satisfy you that ACR *does* add saturation in the manner that I have described. But leaving that aside, I showed in addition to the original, a true luminosity-curve version. While Jeff Schewe is right when he says that certain clients might prefer the garish versions that ACR generated over the true luminosity version, I have jury-tested this image, luminosity curve vs. original, and the vote is 100% for the luminosity curve version. It is surely reasonable that a client might prefer that as a final image and reject anything more saturated as being an unrealistic color for cherries.

So, my question for you is, given the original JPEG, and given the client's explicit instruction to attempt to match the luminosity-curve file that I produced, without increasing saturation, how would you proceed in ACR? If you can't do it, you might wish to rethink your position on how good ACR actually is. That cherries image is bread and butter for a lot of professionals. Any competent program for the last 20 years or so can handle it effectively, including 1980s-vintage drum scanners, Color Studio, Corel PhotoPaint, and most if ot all proprietary raw modules. But not, apparently, ACR.

Dan Margulis
___________________________________________________________________________

Re: [ Color theory and practice
Posted by: Dan Margulis
Mon Jun 11, 2007 5:43 am (PST)

Jeff Schewe writes,

From Bruce Fraser's Real World Camera Raw, page 35, Bruce writes under the
Tone-Mapping Controls section: "The Camera Raw controls use an algorithm
that preserves the original hue, whereas hard curve adjustments to the
composite RGB curve in Photoshop can cause slight hue shifts."
The fact that I believed that paragraph above meant it was a luminance based
adjustment in Camera Raw was my own deduction-not entirely accurate I now
find. And I _LOVE_ to find out when I'm in error because it means I learned
something when I do.

Two quick comments. First, the time I have available for posting to the list is limited, particularly when I'm on the road, as I have been for some time. Posts on this topic have come in faster than I can reply to them. So, I'm paring down the ones that I do reply to, and as to the others, I'm trying to get to them in order of significance. This is particularly true in the case of Jeff's posts--they're full of information. I choose this one because it's the meatiest, but will get to some of the others when time permits.

Second, we would all do well to reread Jeff's gentle and graceful words above. They should serve as an example to us.

Since my original post back when I did my ColorChecker test I HAVE had some
communications with my "bosses" (friends) to find out exactly what Camera
Raw's curves were based on (it's always useful-if you want to know the
truth-to ask the guy that wrote the code...)

[Quoting Thomas Knoll] "I have tried "pure luminance curves" in experiments.
The math is actually simpler and faster than what I'm using actually using,
but the results look bad (to be polite) in most cases.

We all have respect for Thomas and for what he has accomplished during his career. But he is a programmer, not a professional retoucher. If he were, he would realize that there is a difference between looking good and being useful.

If the person using the module is an idiot, then the approach above is a good one--it makes more pictures look better than worse. But if the user is sophisticated, then it's a bad idea. There's no problem getting more color if the picture is too lifeless--ACR itself has some useful tools in this regard. But if the picture is too colorful because of this algorithm, then it may take an expert to fix it, if it can even be fixed at all.

So, the approach above is appropriate for idiots, but wrong for expert users. However, there is no reason to force a single choice down the users' craw. Every other program I've ever seen that's aimed at professionals gives users the CHOICE of how to lighten images.

"It maps the min and max of the RGB values (in linear ProPhoto RGB) through
the tone curve, and the middle value is interpolated to exactly preserve
hue. This results in saturation effects that closely match saturation
effects you get from real film tone curves (which many years of looking at
film photos has taught our eyes to like), while preventing the adverse hue
shifts you would otherwise get from the concave and convex parts of the
curve."

Jeff, this describes a master-curve approach. The luminosity information is being derived *after* the correction. The better approach is to derive the luminosity first and then correct that.

And while you may think you understand my relationship with Adobe...but it's
clear, you don't. Yes, I did consult with Mark Hamburg and Thomas Knoll for
the purpose of improving the sharpening in Camera Raw...but that was Bruce
Fraser's original contract-which sadly, I had to fill in for. I think Camera
Raw 4.1's sharpening is pretty way cool.

I dunno about the sharpening, but I'm rather fond of some of the features in 4.1. The problem is, that, like sharpening, they need to be applied to a full-range image. I hope you can pass on to your friends my congratulations for the effective implementation of Fill Light. I haven't played much with Clarity but it looks good from what I've seen. Vibrance I haven't tested but it should fill a need.

The problem is, these new additions are all icing. There's got to be a cake before they become interesting. Right now, ACR doesn't deliver one.

In this case, the truth is Thomas make an informed decision to do a hue
locked, saturation tweaked, luminance curve adjustment because he thought
that worked better.

It may well--*IN MOST CASES*. But ACR is being promoted as a professional-level tool. "In most cases" doesn't cut it, if it's going to hose the images that don't fall into "in most cases".

There is a lot of terminology being bruited about, but it sure sounds like we can now agree on the following statement: "When one lightens an image in ACR, saturation increases. The operator can not modify or turn off this effect."

If you agree with this statement then I cannot see any possible way of avoiding agreement with the statement I made early in the thread to explain why I had tested the module thoroughly: "I simply could not believe that a major vendor was using a method this crude to open range."

Similarly, if we agree on the first statement--that the command increases saturation, and the operator cannot modify it, then I reiterate my recommendation to those skilled in color correction that it is better to acquire the image in ACR conservatively.

In that role, I welcome learning anything and everything I can as it relates
to Camera Raw...and if _YOU_ have questions regarding _EXACTLY_ how Camera Raw works, you just let me know...I'll pass the questions along. I do have
friends (but no bosses) that are inclined to tell me the truth...and isn't
that ultimately what this is all about...finding out the truth?

Well said. I appreciate the sentiments and believe that the list does as well. Please pass my regards on to your friends.

Dan Margulis
___________________________________________________________________________

Re:Color theory and practice
Posted by: "George Machen"
Mon Jun 11, 2007 12:12 pm (PST)

Jeff, you are absolutely, 100% right! I apologize for wasting your time & that of the rest of the list with my nonsense.

Shortly after I posted my reply, it dawned on me what I had done, and promptly privately emailed the ACT moderator and requested my erroneous message not be posted, but it slipped through the
cracks notwithstanding.*

My convoluted thinking reflected in my comeback represented either a psychotic break, or at best a good ol' early-morning brain fart. Let me be completely clear: I recognize that your explanation points to the ACR tone curve *not* being a composite-channel master-curve adjustment.

(Nevertheless, I still am interested in further inquiry into why Dan's image tests at least partially suggested a master-curve-type result in ACR's tone curve. Something else may be going on.)

- George Machen

--
* I don't blame the ACT moderator. Let me take this opportunity to digress and say that after a number of private correspondences this weekend with John, I came to realize what an unsung, stressful, difficult and time-consuming job the ACT mods endure. These guys also have lives & real jobs and by the end of the day, they are doing their duties frazzled, running on vapor. Allow me to express my gratitude for their voluntary, unselfish efforts!
___________________________________________________________________________

Re: Definition of "Range"
Posted by: Jeff Schewe
Mon Jun 11, 2007 5:43 am (PST)

 "dmargulisnj" wrote:

If there is anything vague about any of these examples, please let me know, or let me
know how many further examples of damage would satisfy you. Since you claim not to
have difficulties with the module yourself, perhaps you would try correcting the images
and seeing if you can get better results. In fact, let me suggest yet another example, which
you will hopefully not find vague:

So, in the examples above, are 1) & 2) based on raw images? (sorry, I don't think I got a copy of PP5E so I don't know)

I will say I saw the gyrations about 3)...but really Dan, that is so far afield as not really useful. Complaining that a JPG'ed image couldn't be corrected because the jpg conversion (from the onboard raw>JPG processor) is "like Camera Raw" is a straw man argument...

Is 4) available as a raw somewhere? Didn't see it in the files or links (unless it's called something weird).

6), I thought Mike and I already pointed out that the Camera Raw curve version matched neither a Lab nor Photoshop composite curve adjustment...and that was starting from a scan turned into an sRGB jpg...

So, my question for you is, given the original JPEG, and given the client's explicit
instruction to attempt to match the luminosity-curve file that I produced, without
increasing saturation, how would you proceed in ACR?

"original JPEG" + "match the luminosity-curve" + "proceed in ACR" = none starter...

Dan, don't know if you noticed or not but Camera Raw doesn't have ALL the stuff that Photoshop has...see, you START in Camera Raw and you OPEN in Photoshop...you can't do localized correction in Camera Raw either...yet. There's a lot you can't do in Camera Raw that you can do in Photoshop. So, how do YOU suggest Camera Raw be used? Ignore the controls and settings, and process a null conversion into 8 bit sRGB for correcting in Photoshop?

I don't find that particularly useful...that's what I did back in 2001 when I FIRST started shooting digital. I used Canon's really, really bad software to convert the raws to 16 bit Adobe RGB tiffs...and tossed the raws (and boy am I sorry). Compared to then, working in Camera Raw (particularly 4.1) is a tremendous time saver that allows a skilled operator to process raw images to a very high degree of quality. Superior to film.

Do you find otherwise?

Jeff Schewe
___________________________________________________________________________

Re: Definition of "Range"
Posted by: J Walton
Mon Jun 11, 2007 5:44 am (PST)

On 6/10/07, dmargulisnj wrote:

5) Same picture. Members of this list posted three or four examples that the photographer (and I, and everyone else who commented on them) thought were good renditions. Claim: that none of these straightforward global moves could have been achieved in ACR by any combination of commands.

I don't want to cause any problems, and I definitely don't want to get in the middle of the current "discussion" but I should interject something as it pertains to the ballerina image I personally submitted.

There's now way I would have achieved the results I did by working from the JPEG - I had to go from the RAW file. There is also no way I would have achieved those results in half an hour *without* manipulating the image in ACR, and probably wouldn't have ever made it quite there.

Realistically, the temperature correction in Camera RAW (CS2) got me 1/3 of the way there. It changed the hue and kept the colors separate enough that I could then *start* to work.

I don't know whose side that helps or hurts, and I don't really care that much, but that's the truth. As usual, intelligently using all the tools available resulted in the best possible image. The real challenge is in deciding which tools are best suited to which task. I suppose that is the constant struggle that is made both easier and more difficult by the abundance of ever-changing tools.

--
J Walton
___________________________________________________________________________

Re: Definition of "Range"
Posted by: "Mike Russell"
Mon Jun 11, 2007 12:12 pm (PST)

From: "Jeff Schewe"
...
6), I thought Mike and I already pointed out that the Camera Raw curve version matched neither a Lab nor Photoshop composite curve adjustment...and that was
starting from a scan turned into an sRGB jpg...
...
As long as I'm being quoted here, I'll repeat "the rest of the story", which is that I have verified to my own satisfaction that the Camera Raw curve is not a composite curve, but that it modifies saturation, and that certain hue values change significantly as well.

These saturation and hue changes substantiate Dan's broader point that the Camera Raw curve is unsuitable for professional color correction. When curve access to all the channels is quietly added to ACR, Dan will get a large share of the credit, in my book.

The exact math behind the curve, while interesting, is not germane to the broader point, any more than the exact source of a color cast matters so much as the need to remove it. Seizing on it is like interrupting a lecturer for small grammatical errors, when one has not bothered to read the assignment.

Mike Russell - www.curvemeister.com
___________________________________________________________________________

Re: Definition of "Range"
Posted by: Dan Margulis
Mon Jun 11, 2007 5:50 pm (PST)

Lee Clawson writes,

But I'm still confused. I know what it means to set a range but what are
"range-opening routines"

When we "set" a range it is almost certain that we are making it larger than it previously was. The only times we make the range smaller is when the object is not significant, and in that case we rarely refer to its range. "Opening the range", "extending the range", "expanding the range" "setting the range," "establishing the range" are all common ways of describing the process of setting highlight and shadow, or of emphasizing a certain object.

and why are they inferior in the module ??

Automated color correction algorithms like those in JPEG assume that the user is not sophisticated. They correctly assume that the users wants bright, vivid, pleasing results most of the time. They therefore emphasize extended range and saturated colors. They wouldn't do this if it wasn't correct to do so most of the time, but unfortunately there are many occasions when it is *not* correct. When that is so, and the JPEG does it anyway, the image can become difficult to impossible to fix.

The major reason for the existence of raw modules is the ability to bypass this automated intelligence when necessary. While ACR lets us capture a flat-looking image that doesn't have these defects, we then have to correct it in Photoshop. We'd like to be able to establish a full range in ACR, but we can't, because it uses much the same "you are not a sophisticated user, therefore you want bright colors" approach that the JPEG algorithms do.

Again, we do want these brighter colors most of the time. But we don't want them *all* the time. Every other professional-level program gives us several options of how to establish range. But with ACR, you are getting more saturated colors, whether you like it or not--no way to bypass it. That's not acceptable for a professional user, and that's why the routine is inferior.

Is there something you prefer that's much better ??

There are lots of ways that are much better. That's the good thing about having the raw data--we can dress up the interface however we like, with whatever kind of artificial channel structure and other bells and whistles we can think up. The simple way, which many other modules use, is to graft on an RGB curves interface. That would work for ACR--it would allow range to be set properly. But it is not necessarily the best way, let alone the only one.

Much depends on a question that Jeff Schewe asks elsewhere: what do you want Camera Raw to be?

It seems to me that there are five basic image-fixing workflows that involve a raw module. From simplest to most complex, they are:

1) All we are trying to do is get the data in a form that we can correct it, without loss of anything. We don't care what the subject of the image is or where it came from. We just acquire in the simplest reasonable way.

2) We have gone to the trouble of making some specific settings that pertain to the type of images that we shoot, or to images that come from some other source, are of some specific subject, etc. When such images show up, we open them using the appropriate settings that we have developed, but we don't usually do anything image-specific--if the image is of a certain category, we open it a certain way.

3) We are interested in good but not great quality. We will correct images individually in the raw module to make them look as best we can, but that's going to be it. It doesn't matter if we could do more in Photoshop afterwards--we're just not going to do it, because the extra quality isn't worth the time.

4) We *are* interested in good quality, and we assume that Photoshop is where we will get it. However, when we open in the raw module, we are willing to take a hack at it to get it closer to a good starting point for Photoshop. IOW, if the image is obviously too green, we take 15 seconds to push it somewhat in the direction of where we want it to go eventually. We are not trying for perfection, we just think that the 15 seconds we spend now might save three or four minutes later.

5) We intend to take all the time needed to make the best possible image in the raw module, and we reserve the right to make further changes in Photoshop as necessary.

As matters stand, ACR performs well for #1 and #2. For #3, it's OK but not very good. It needs some means of applying a correction that augments detail in a single object more precisely than a master adjustment can. It could be channel curves, or it could allow the user to break out the detail in some more novel way, but the current offering isn't adequate. As against that, ACR now has an interesting lot of addons that the competition doesn't. If you compare it to, say, Canon's DPP, it's hard to say which is better from the standpoint of workflow #3.

For workflow #4, it's not hard at all. ACR is just plain non-competitive. With the mandatory increase in saturation, we can't guarantee that our correction won't damage the image. That could be fixed by adding RGB channel curves, too, but one should recognize that RGB is the worst space for adding contrast. It would be better to offer at least one other colorspace--LAB, CMYK, or even HSB--as an alternate.

Workflow #5 is, I think, what Adobe and the camera vendors would like us to adopt, but it's a *really* hard sell. Few people want to spend a lot of time correcting in a module and then a lot of time in Photoshop, when the same thing could have been done in Photoshop alone. To make it viable, one of two things would have to happen:

a) the raw modules could become so powerful that we would rarely have to enter Photoshop at all, except for such things as local retouching.

b) the raw modules would have to be WAY better than Photoshop at certain operations--like, for example, in setting initial ranges.

I don't think #5a is possible for at least three more releases--too many missing pieces. #5b IS possible but would entail work. One would want to have all the usual options for range enhancement but there would have to be considerably more to make the workflow attractive. We are beginning to know enough about color to start to think in terms of more sophisticated, artificial ways of setting the range.

For example, I'm playing around with what might be described as a five-channel artificial colorspace, using light red, light green, light blue, gray, and black. I'm not saying that's the best way to do it, but it's quite powerful. If a few such beasts were grafted onto the interface as an option, and if the white balance commands were upgraded with a decent range slider, then you *would* have something that's way better than Photoshop at acquiring an image.

But don't hold your breath. For the time being, I think we should simply hope that, as Mike Russell suggested, channel curves get slipped into the next release.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: J Walton
Mon Jun 11, 2007 5:44 am (PST)

On 6/10/07, Mark Segal wrote:

What was known by scanner operators is not necessarily
relevant. We're dealing with fundamentally different
technology and image processing algorithms using
mathematical techniques some of which were quite possibly
not yet invented in those days.

I see where you are coming from but disagree in that the scanner operator and the image processor are both addressing the same basic task. What was known by the best scanner operators of 20 years ago can help with image capture and conversion today. IMO, it's relevant.

But I have no further interest in that element of the
discussion because I wasn't a pre-press scanner operator -
I'm speaking from deductive logic rather than the actual
experience of comparing luminosity adjustments of a similar
photograph scanned for pre-press versus processing a digital
file of the same type of image in Adobe ACR 4 and CS3. Have
you done that comparison so you know that all the
constraints and impacts of scanning apply pari-passu to
digital imaging with today's applications?

I was a pre-press scanner operator and I'm just now getting used to ACR. Up until a couple of years ago I'd never used it even once. I'm finding what you can do with a RAW file is quite impressive, but I'm also finding that my scanning experience helps me to use ACR. I'm not sure what pari-passu means but I do see the correlation between scanning and ACR.

Where I do have much actual experience, and thinking also
about conversations with other very good digital image
makers, I know that using the RGB composite curve is
harmless on a very wide range of full colour photographs
regardless of what those B&W channel grabs of cats show you
in PP5E. Yes, there are images where individual channel
adjustments may be necessary - I'm not debating that. But
instead of just reading and quoting the book, have you tried
both techniques yourself on a fairly large random selection
of real world photographs, printed them in full colour in a
professional inkjet printer and looked at them? If you do
that, I'd be mildly surprised if you didn't find that I may
have a point.

I think you have a very good point. I know that Dan has cracked pretty hard on master curve adjustments, and I have learned a lot from him over the years, but I just don't see eye to eye on the dogmatic master curve argument. When a curve move needs to be done I probably use a master curve 40% of the time, a luminosity or color move 40% of the time, and individual channels 20% of the time. And that suits me just fine because I know the kind of images that a master curve would damage and don't even attempt it on those. In other cases I *simultaneously* want to increase contrast and increase saturation at the *same time,* and I can do that in one move rather than several.

Often for me I'm looking for a particular "look." I start with a master channel and sometimes I say "Wow! That's looks great" and move on to another correction. And I don't feel guilty about it either.

--
J Walton
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Ric Cohn"
Mon Jun 11, 2007 12:12 pm (PST)

On Jun 8, 2007, at 7:02 PM, Jeff Schewe wrote:

I'm working on the rest of the response to this post, but I can tell you that doing it Dan's
way, bringing the untagged and presumed sRGB jpg image into Camera Raw 4.1, it sure
looks ALOT more like the Lab corrected image than the Photoshop composite RGB curve
adjusted image...unless I'm doing something wrong.

I used Lightroom on this image and that hasn't had 4.1 posted yet so It's "only" 4.0. Unless there is a change in this behavior (or I'm doing something wrong) I'm seeing exactly what Dan explains. If this is a 4.0 to 4.1 improvement I can't wait to see what I would consider a welcome improvement. I opened the jpeg in Lightroom and tried both increasing the contrast with the contrast slider and using the master Curve. Both looked very similar to the image where I used a master curve in Photoshop. With greatly increased saturation along with contrast. I also tried converting the sRGB original to ProPhoto RGB to see if Lightroom handled it any differently. It didn't appear to change the results.

As has been pointed out many times by Dan and others, usually we are happy about an increase in saturation, and indeed the cherries *look* better to me with a master curve. As we all know, this on screen appearance can hide a lot of file damage that is difficult to correct later.

I've always believed in getting the best image out of the raw without doing any damage that might only show up in later corrections. As raw processors have improved I've been willing to do more fine tuning with each advancement. I find those arguing that ALL the "global" corrections should be done in ACR or Lightroom (or any other raw processor I've worked with) and the image should be suitable to final output out of the Raw processor unconvincing- at least for my own images. For one thing, I do not believe the tools are there to evaluate the image properly for damage. I don't believe the Histogram is enough. The absence of an option to look at numbers in anything but RGB, the absence of Soft Proofing, and (unless I've missed it) the inability to view the individual channels are just 3 limitations which I hope will be added soon.

Ric Cohn
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Mon Jun 11, 2007 12:15 pm (PST)

J Walton, I did say yesterday that I was abstracting myself from this debate until I had more substative things to say about "testing", but in light of your comment below on scanning and a similar one I received off-line from another party, I feel I need to clarify what I meant and it is this. Indeed, there are basic, fundamental things, (e.g. the nature of the image corrections that need to be made, the core of how curves behave - e.g. steepening them increases contrast, the nature of the colour wheel) that carry from one technology to another. But the technology for implementing tonality and colour adjustments and the impacts of that implementation on the images must have changed very dramatically over time because of the explosion of powerful new techniques - some of which depend on math that didn't exist some years ago, from what I have been reliably told. That is why I find it suspect to draw inferences about the behaviour of curves in ACR from what one knows about pre-press scanning of yesteryear. It's not that there's anything to belittle about pre-press scanning - it's just that the world has moved-on and "the old dogs have gotten new tricks" that work very well. All of this was in aid of supporting an argument that it makes more sense to be pragmatic about these issues, very much as you express below.

Mark Segal
___________________________________________________________________________

Re: Definition of "Range"
Posted by: Dan Margulis
Tue Jun 12, 2007 5:23 pm (PST)

J writes,

I don't want to cause any problems, and I definitely don't want to get
in the middle of the current "discussion" but I should interject
something as it pertains to the ballerina image I personally
submitted. There's now way I would have achieved the results I did by working from the JPEG - I had to go from the RAW file.

I don't know whether it's possible to get that good a result from the JPEG but it would certainly be harder. I stated right up front that this image was an excellent example of both the strengths and weaknesses of ACR. Strength, in that the damage caused during the JPEGging could be reversed. Weakness, in that the range can't be set properly in ACR in this type of image without provoking similar damage.

I stated, before anybody else had worked on the image, that anybody who opened the range in ACR would probably have to resort to a channel blend (and/or, it went without saying, local retouching) to fix the color. Your successful version, as predicted, used both channel blending and retouching.

Realistically, the temperature correction in Camera RAW (CS2) got me
1/3 of the way there. It changed the hue and kept the colors separate
enough that I could then *start* to work.

Using that tool makes perfect sense. The thing that concerns me is that this combination (set range and white balance in ACR, channel blend) is likely to deprive the face of any real-looking color variation, even though the face itself may be the desired color. I'm thinking of feeding this image to the advanced classes in November to find out the answer.

I don't know whose side that helps or hurts, and I don't really care
that much, but that's the truth. As usual, intelligently using all the
tools available resulted in the best possible image. The real
challenge is in deciding which tools are best suited to which task.

As always. I agree, of course, with all the other observers that your image is excellent.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Wed Jun 13, 2007 10:55 am (PST)

Jeff Schewe writes,

So, what did you do to the file while it was in Photoshop CS2? Pretty sure
Camera Raw 4.1 only works in CS3. So, did you open it in CS3, save and
then reopen in CS2?

It appears from the rest of your post that you have confirmed that the files are accurate, in spite of the fact that they were saved out from CS2. Nevertheless, since the story both answers some questions from other members of the list and offers a feature suggestion that your friends may be interested in, here goes.

As noted in an earlier post, when a Photoshop upgrade occurs, I try to have both the new and old versions loaded on each student's system. As color correction technique varies little from version to version, students are permitted to use either version for their exercises. However, I only teach in the new version in classes in which 2/3 of the students say they've switched themselves. The sponsor, by contrast, names a date certain upon which the change is made. For those interested, CS2 is Ledet's default for teaching until Friday; on Monday CS3 becomes the default. ACT is the exception and will continue to be taught as described above.

In the class that began Thursday, only two of the seven students had shifted to CS3. In the one the previous week, in response to a list query, I noted that only one student used CS3. These figures are normal, in my experience, for Photoshop upgrades.

Therefore, while my machine had brand-new installs of both CS2 and CS3, I personally was using CS2, *except* to produce the two test images, for which CS3, as you noted, is required. Having saved out the two new JPEGs, I closed CS3 and reverted to CS2.

Being somewhat pressed for time, I will confess to thinking that when one opens a JPEG in ACR, and then resaves it out with a name that includes a .jpg suffix, the module might conclude that it should be saved in JPEG format. This is, regrettably, not the case in CS3. My files were in PSD format, as I discovered later when I attempted to open them for a final check before posting. I therefore renamed them with a .psd suffix, opened them in CS2 because that's what I was using at the time, and resaved them without change. I have the original CS3 PSDs if you like, but I am sure you realize that they will show the same thing.

My suggestion for your friends is that it would have been helpful if, when saving out of CS3, I had received an error message stating, "your chosen format does not agree with the suffix you have chosen. If you save the file this way, Photoshop won't be able to reopen it unless you rename it."

In the grand scheme of things, using Camera Raw on jpgs, is, well kinda an
odd way of using Camera Raw...but, whatever.

Recall, please, that I urged people to use their own images and described the sorts of image that would show the effect prominently. I only posted the JPEG for those who did not have such an image at hand. I used a JPEG rather than a raw because the smugmug site does not support raw images. There is a similar raw image of a red flower on the PP5E CD, and the book shows similar damage when the range is established in ACR.

However, I did process your image with CR's curve using 64=45 and 195=215. Then processed the original jpg with linear (no curve) and applied the same curve points in Photoshop. The Camera Raw curves does not match the appearance of the Photoshop curve version.

Nobody said that it would. I said that it would have certain characteristics that would identify it as being a master-curve type approach and that these characteristics would easily be seen to be shared by the Photoshop master curve version.

I also took the original jpg, converted to Lab and tried to match the curve
(course I can't use the exact same input/output numbers since Lab ain't 0-255)
and looked at all three images:
Lab curve: (input: 24, output: 16 & input: 75, output: 87) color looks flat and
a bit unnatural but the contrast increase had zero effect on saturation...

The effect on saturation is not zero, but it's negligible. I agree with your comment about the color (and that we can't match the curve exactly due to gamma variations, among other things), but we need to make one further observation. Comparing the cherries after the true luminosity curve to the original, there is an unambiguous contrast gain EVERYWHERE in the fruit. The sides of the cherries, which are originally relatively dark, have gotten darker. The pink reflections have gotten lighter or stayed the same. The drops of water are incontestably better defined against the fruit.

Camera Raw curve 64=45 and 195=215 increased the contrast and has an
effect, mild, on saturation...

I can't agree fully with either statement and invite you and others to revisit the pictures yourself to see whether you concur with the comments below. For those interested in the actual images, the original, luminosity curve, and Photoshop master curve version are at:
http://colortheory.smugmug.com/gallery/2971728

and the version produced by ACR's curve is at (Jeff and I are speaking about the duller, less radical of the two that are posted):
http://colortheory.smugmug.com/gallery/2973533

a) What you are taking for better contrast is a darker shadow--the dark areas between the cherries have gotten darker. This behavior is consistent either with a luminosity curve or a master-curve approach. However, unlike the real luminosity curve, the ACR version shows loss of detail in the lighter parts of the fruit. In the top left cherry, look at the slight reflection that covers most of the left side. Here, the ACR version has *less* contrast than the original, the reverse of the luminosity version, which has more. As this area falls (in terms of luminance) into the area in which we are applying a steeper curve, it is impossible that it should be losing contrast if a true luminosity curve is being applied. Instead, it is visually similar to the Photoshop master-curve version, which also loses contrast there.

b) With regard to the "mild" effect on saturation, I've chosen four points to compare, going from darkest to lightest. These are all LAB equivalents although the files themselves are all sRGB.

i) Left edge of the center cherry. 29L45a34b in the original, 24L44a34b in the luminosity version, 25L44a36b in the ACR version. Granted that we have asked for a darkening of the area, which accounts for the change in L values, what ACR does here is reasonable.

ii) Center of the center cherry. 40L51a42b in the original, 38L52a42b in the luminosity version, 39L56a52b in the ACR version. This large saturation gain drives the color out of the CMYK gamut and also that of many inkjet printers. Also, there is a hue shift toward orange.

iii) Center of the right-hand cherry (the one with the visible stem). This color is originally more orange than in the other cherries. 45L51a48b in the original, 44a52a48b in the luminosity version, 46a58a59b in the ACR version. This is not a mild, but a *massive* increase in saturation, as well as a continuation of the hue shift toward orange. That the saturation boost gets larger quickly as the tone gets lighter is characteristic of the master-curve approach. It is inconsistent with a luminosity-based approach.

iv) The aforementioned light area on the left side of the top left cherry. The original and the luminosity version are identical at 46L39a9b, or, in sRGB, 169r78g94b. The ACR version is 42L62a19b or 181r50g81b. This increase in saturation is even greater than in iii) if possible, and this time the hue shift is toward purple. As noted before, that the ACR version got darker is consistent with a master curve and inconsistent with a luminosity-based approach. Also, that certain channels are getting drastically lighter and others drastically darker is consistent with a master curve and inconsistent with a luminosity-based approach.

These numbers show that several suggestions that have been offered are either not correct or don't work, namely:

1) The suggestion that one could eliminate this saturation gain by adjusting some other control in ACR does not work, because the saturation gain varies profoundly with the darkness of the affected area, and no ACR command has the adjustable masking capability that would be needed. The white balance routine *is* darkness-related, but changing it would affect parts of the image that are currently correct.

2) The suggestion that there is a "hue lock" on this method is refuted by the above results, which show that the hue doesn't just change, but it changes in different directions at different darkness levels.

3) The idea that this is a "mild" saturation boost is refuted by what is happening to the lightest areas above. If asked to describe the colors to a layperson, I'd say that in iv) above the original color is a subtle dark pink. I asked for a contrast boost in ACR and the color became what I would describe as an iridescent, brilliant red, with loss of detail.

Which is "better"? Well, that is certainly open for debate. But what is NOT
open for debate is the statement that Camera Raw uses a "Master Curve"
approach that is the same as Photoshop (at least in CS3).

To repeat, nobody said this. You quoted the words that I *did* say in one of your own messages. "This gallery is an aid to those wishing to test whether the Curve function in Camera Raw uses as its model a master curve based on RGB-type channels, or whether it is based purely on the luminosity of the image". This statement does not imply that you would get an identical result to a Photoshop RGB master curve, or, for that matter, the curve function of any other program.

As to which way is better, nobody knows this except the client. Maybe they want the brilliant red that I didn't ask for, but perhaps they don't, and if they don't, this file is damaged goods. Certainly we *want* to be able to establish brighter reds. It can even be the default setting. But to be *forced* to accept a color change of this magnitude is not practicable for a professional.

You do not have to be a numbers person to explain the ramifications of this to your friends. Nobody needs terms like "master curve", or "luminance", or "hue lock". You just have to say that there can be a really big change in color when this command is used, and a loss of detail in brightly colored areas. You say that the magnitude of this change is big enough that some people won't set range in ACR because of it. And, you point out what should be clear: there have been what look like really good features added to 4.1, but they all depend upon establishing a proper tonal range.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Thu Jun 14, 2007 2:36 pm (PST)

Dan Margulis wrote:

Being somewhat pressed for time, I will confess to thinking that when one
opens a JPEG in ACR, and then resaves it out with a name that includes a .jpg
suffix, the module might conclude that it should be saved in JPEG format. This
is, regrettably, not the case. My files were in PSD format, as I discovered
later when I attempted to open them for a final check before posting. I
therefore renamed them with a .psd suffix, opened them in CS2 because that's
what I was using at the time, and resaved them without change. I have the
original CS3 PSDs if you like, but I am sure you realize that they will show
the same thing.

I suspected this might be the case...but thought it was useful to point out that there were two different Photoshop versions at play...I didn't really think you were trying to get away with anything.

My suggestion for your friends is that it would have been helpful if, when
saving out of CS3, I had received an error message stating, "your chosen
format does not agree with the suffix you have chosen. If you save the file
this way, Photoshop won't be able to reopen it unless you rename it."

Well, I suppose it would be useful to have Photoshop keep people from making errors...but then again there's a lot of squealing already with Photoshop appending the word 'Copy' when saving out with or without layers-depending on the current state-so, I suspect you can't please all the people all the time, ya know?

I can't agree fully with either statement and invite you and others to revisit
the pictures yourself to see whether you concur with the comments below. For
those interested in the actual images, the original, luminosity curve, and
Photoshop master curve version are at:
http://colortheory.smugmug.com/gallery/2971728

and the version produced by ACR's curve is at (Jeff and I are speaking about
the duller, less radical of the two that are posted):
http://colortheory.smugmug.com/gallery/2973533

Well, I see your Luminance Curve and raise you one Camera Raw adjustment that comes REAL CLOSE to your Luminance only curve done in Lab...

I've taken the liberty (with Dan's permission to access and use the original sRGB jpg image) my whack at producing a contrast adjustment that pretty much matches your Luminance curve adjustment, done all inside of Camera Raw 4.1.

See the gallery at:
http://colortheory.smugmug.com/gallery/2999719

The first image is the original sRGB image in it's original lower contrast version. The second image is the one Dan adjusted with a Luminance Only curve in Lab then converted back to sRGB. The third image is my attempt to match Dan's curve inside of Camera Raw 4.1. You can download the Camera Raw 4.1 settings (you will need at least Camera Raw 4.0 to use them). The preset is available from:

http://tech.groups.yahoo.com/group/colortheory/files/
and the file name is: 'Cherries-Luminance.xmp.zip'

Is this an ABSOLUTE EXACT MATCH for a luminance curve, no...but it doesn't have to be EXACT. All I was trying to do was prove that Camera Raw's current tool set allows a lot of flexibility to make all sorts of adjustments including adjusting the contrast while also backing off the saturation boost, if necessary. Yes, I adjusted the master saturation as well as the HSL saturation for red, orange and yellow as well as color luminance. How long did it take? Well, I popped Dan's Luminance jpg and the original both into Camera Raw in filmstrip mode so I could toggle back and forth to see when they match...and yes, I did twiddle (I wanted to get pretty close ya know). Took about 3-4 minutes.

The worse part of the exercise was the fact the original image I saved out with the luminance image was actually saved out as ProPhoto RGB and uploaded. -DOH- The image is there as neither Dan nor I can seem to delete it. It's the forth image in the gallery and it's labeled as Oooops, PP RGB as it would look in sRGB. It's at least educational...just don't let it confuse the issue at hand...luminance type curves in Camera Raw.

And this is indeed a pretty important topic. And, I've tried to reconcile Dan's contention that Camera Raw is not a professional level tool with my perception that's it's a really great & valuable tool for digital photographers...how could our perceptions be so far apart?

Then a realized something...Dan isn't a photographer. His impressive career and experience and expertise is based primarily on taking film-something that already fixed the original scene contrast-and correcting problems (often caused by poor photographic technique).

My whole photographic career has been fighting the battle of getting the scene-often far wider than can be captured-first on film and then later on digital capture. So, what Dan doesn't really understand is that in the grand scheme of things, the battle of capturing the original scene is far more likely to involve _REDUCING_ the scene contrast to fit into the more limited range of film or sensor, not increase it-particularly when you factor in final output which has an even more reduced dynamic range.

And, as a direct result of that main problem-too much scene contrast-Camera Raw has been designed, first and foremost, to excel in range reduction, not increase...hence the attention to things such as Fill Light and Recovery.

Those times when I, as a photographer, encounter really, really flat light and actually NEED to increase the range (contrast) a slight boost in saturation is not only not bad but desirable. Which is why Thomas says he decided NOT to use a luminance only curve but one that would more closely resemble film being increased in contrast. He did it on purpose.

Dan's point of view, coming from where he is coming from is understandable (even if he's not a photographer)...he's comparing Camera Raw to tools he's used and is familiar with...and so am I. I've tried all the raw converters out there and I seriously like Camera Raw the best-not because I played a tiny role in it and consider Thomas a friend, but because it produces really good results when working from raw captures to color space encoded images in Photoshop. So what if the saturation bumps up when I increase contrast...cool! It just means that there's less saturation I need to boost for those low contrast scenes I shoot. Remember Kodachome 25 and Velvia? It did what to colors? Punched them up?

Can Camera Raw do curve increase without doing a saturation boost? Nope...but if you really, really want to reduce saturation, you can sure come close to a luminance only curve adjustment-see the gallery above. (the settings needed will vary by image).

So, the perspective regarding Camera Raw's professional creds really comes down to where you are coming from. As a photographer, I and others like the result that Camera Raw makes possible-including simulating a luminance curve only correction for those rare times when that's what you need. From the perspective of graphic arts and prepress production, maybe Camera Raw isn't the tool for you. Course, we are fortunate that we then can rely upon Photoshop to do those kind of corrections-and isn't it convenient that when you open an image from Camera Raw, it opens into Photoshop?

So, what have we learned? Camera Raw doesn't use a Master Channel adjustment like Photoshop, it's a hybrid that Thomas tuned for working with raw images. Can you do contrast increases with little or now boost in saturation, yes, but not with a contrast curve only. You'll need to twiddle some other sliders. The vast majority of use cases has Camera Raw working on wider dynamic scenes where contrast (or range) reduction is more important that contrast increase. And for photographers (at least THIS photographer) Camera Raw is a nifty professional level raw image adjustment tool...but, it's ain't Photoshop.

...and, neither Dan nor I can figure out how the heck to remove an image from SmugMug once the darn thing has been uploaded...

:~)

Regards,
Jeff Schewe
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Ray Van Dusen"
Thu Jun 14, 2007 7:12 pm (PST)

Jeff Schewe said:

My whole photographic career has been fighting the battle of getting the
scene-often far wider than can be captured-first on film and then later on
digital capture. So, what Dan doesn't really understand is that in the grand
scheme of things, the battle of capturing the original scene is far more
likely to involve _REDUCING_ the scene contrast to fit into the more limited
range of film or sensor, not increase it-particularly when you factor in
final output which has an even more reduced dynamic range.

And, as a direct result of that main problem-too much scene contrast-Camera
Raw has been designed, first and foremost, to excel in range reduction, not
increase...hence the attention to things such as Fill Light and Recovery.

I've missed reading the contextual emails that this one arises from but felt that I had to confirm Jeff's experiences and conclusions here. As a long time photographer as well as photo teacher, I and my students have spent much energy in taming the higher contrast inherent in (the majority of) our scenes.

In B&W film days it was always "expose for the shadows and develop for the highlights". When we shot chromes, we learned to tame contrast with the use of variations of fill light. Now with Photoshop, thank god we have ACR's "fill light" and "recovery" (why not just expand the latter to "highlight recovery" BTW?) not to mention "Shadow and Highlight" in Photoshop itself.

Of course we use Curves and masks, etc. for both tweaking and substantial adjustments but these aforementioned features have been a godsend in dealing quickly and efficiently with some aspects of the workflow. It's no mystery to me that these features have been designed specifically with the photographer in mind and I and many photographers I know use them all regularly and gratefully. (Thank you Thomas!-)

- Ray Van Dusen
--
Ray Van Dusen, MFA
faculty
Professional Photography Dept.
Dawson College
http://dc37.dawsoncollege.qc.ca/photo/ray_v.html
http: //theworkshops.com/catalog/faculty/index.asp?SchoolID=20&FacultyID=885
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Thu Jun 14, 2007 9:43 pm (PST)

Ray, yes, exactly. Every time I rescue an image with Recovery, Fill Light, and Blacks I mentally thank Thomas as well. They are very effective tools. The new Vibrancy, HSL Tab and Greyscale functions are remarkable as well.

As you say you have not been immersed in all of this thread (perhaps mercifully), let me bring you rapidly into the picture. The fundamental gist of the discussion boils down to whether or not increasing contrast by using the Curves in ACR (which are complementary to the above-mentioned controls) produces a large enough saturation boost to smother image detail, to what extent ACR's controls are adequate for rolling-back excessive saturation (which may not occur in all parts of the tone curve the same way), and from there whether this danger seriously impairs the usefulness of the application for professional purposes. The debate really centers around whether these are major/minor, fatal/non-fatal issues.

In evaluating this debate, apart from the evidence of one's own work starting from raw files, one should be mindful of the fact that the saturation boost of interest here was programmed into the application on purpose for aesthetic reasons - i.e. photographic taste. So then one can argue about whether or not this is a *good thing*. I've been playing with that issue in Photoshop by increasing contrast as a pure Luminosity function or the *Normal* way. More often than not, the added saturation looks more natural. Where it doesn't, it can usually be fixed if one didn't do irretrievable damage to the image beforehand. One thing that's nice about all this in ACR/Photoshop is that unless you really don't know what you're doing, it is possible to do a lot of destructive testing to see what happens under what conditions, and then recover from it.

Turning to the evidence of one's own work, I can only speak for myself - like so much else in digital photography, I happen to believe the answer is "it depends" - I know, not very determinative, but that's both *on purpose* and I think *appropriate* to the nature of the question, where there are numerous changing variables at play from one image to the next. I also happen to believe, based on about a thousand images processed with ACR 4 (beta or actual) over the past few months, that judicious use of this application generally produces very satisfactory results, without however obviating some important post-rendering adjustments for which Photoshop remains necessary (e.g. final image adjustments under soft-proofing, or for extreme luminosity problems - the kind of masking and luminosity blending for which Photoshop is so well adapted, etc.).

Mark Segal
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Ron Kelly"
Thu Jun 14, 2007 9:43 pm (PST)

Gentlemen:

I think you are misunderstanding something here.

As a photographer I understand that yes, often one is trying to capture a scene of higher contrast than the process can readily accommodate in a single step.

However, whatever methods used be they fill lights, bracketed exposures blended, scanned film and fancy channel blends, etc. you have an image file that is imperfect for output after all that.

At that point you're going to expand the range for output.

A carpenter has a board that's too short. He glues two or more boards together, and he still cuts some off to make it the _PERFECT_ length.

Ron Kelly
___________________________________________________________________________

Re: Color theory and practice
Posted by: "John Arnold"
Thu Jun 14, 2007 9:52 pm (PST)

Jeff,

Very interesting post. I would agree, it's not exact. But it's close. As one user mentioned earlier, this dialog has been very educational for those of us interested in learning more about Camera Raw 4x. It's interesting to see how fairly similar results can be achieved in Camera Raw 4x and Photoshop, provided of course that one accepts the fact that you have to travel a different path in each program to obtain them.
 
I have a question that I posted a couple of weeks ago that is a bit off-topic here. But since you had some input in the development of the sharpening adjustments in Camera Raw 4.1, I thought I would ask you if you could answer it. In Bruce Fraser's sharpening book he suggests protecting the extreme ends of the tonal range by using the blend-if sliders when doing your capture sharpening. Do any of the new sharpening controls in Camera Raw emulate this technique? By the way, I did read the tutorial you posted and it was quite helpful.

Stephen Marsh suggested that I do a comparison by using the sharpening controls in Camera Raw and overlaying the result in difference mode over my normal capture sharpening process in which I do use the blend-if sliders to protect the ends of the tonal range. The differences were almost non-existent. So I suspected that there might be some similar dampening for the endpoints built into the Camera Raw controls. I'd be interested to hear your comments.

Thanks,

John Arnold
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Fri Jun 15, 2007 3:34 am (PST)

Ron Kelly wrote:

At that point you're going to expand the range for output.

Well, I don't know what printer _YOU_ use, but it's my experience that I rarely if ever need to expand the range for output...it's usually required that I further reduce the rangedue to the reduced dynamic range of prints...

Unless I'm not understanding what you are saying...

It's my experience that deep shadows block up which requires lightening (reducing range) and highlights that require darkening not to burn out (again, range reducing). There are times that I may need a mid tone lighening...I guess that could be called range expanding...

I usually can tell this from soft proofing...but then again, this is done in Photoshop...not Camera Raw.

Naw, the vast majority of images I have to deal with involve reaching into shadows and pulling back highlights...and those fall under the heading of range reductions, not expansion.

Jeff Schewe
___________________________________________________________________________

Re: Color theory and practice Now sharpening in Camera Raw
Posted by: Jeff Schewe
Fri Jun 15, 2007 3:36 am (PST)

 "John Arnold" wrote:

In Bruce Fraser's sharpening book he suggests protecting the extreme ends of
the tonal range by using the blend-if sliders when doing your capture
sharpening. Do any of the new sharpening controls in Camera Raw emulate this
technique?

But of course...when I was at Thomas' house working with he and Mark Hamburg on the Camera Raw 4.1/Lightroom 1.1 sharpening, Mark had Bruce's book out inspecting and evaluating Bruce's blend options...I'm pretty sure they aren't exactly the same however because the sharpening is being done in linear gamma instead of a 1.8 or 2.2 gamma of a regular color space.

But yes, the sharpening effects are tapered off the shadows and also off the textural highlights to avoid slamming any texture to pure pinned white.

It should also be noted that the Detail function is a medium frequency pure halo suppression technique Mark devised to also help increase the midtone sharpening effect while mitigating shadows & highlights....

Bruce had a big impact on the sharpening, no doubt about it. It was Thomas himself that wanted Bruce.

Regards,
Jeff Schewe
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Ric Cohn
Fri Jun 15, 2007 6:46 am (PST)

On Jun 14, 2007, at 5:09 PM, Jeff Schewe wrote:

Well, I suppose it would be useful to have Photoshop keep people from making
errors...but then again there's a lot of squealing already with Photoshop
appending the word 'Copy' when saving out with or without layers-depending
on the current state-so, I suspect you can't please all the people all the
time, ya know?

I've made the same dumb error and I was actually hoping for a warning in CS3. I think that unlike the "Copy" controversy there is no question that if you do something that appends the wrong file type you've done something you didn't mean to do. I've done this when basing a new name on another file in the "Save" folder and I click on a file with the wrong suffix and forget to change it before hitting Save. Careless, yes, but one of those times I appreciate if a program stops me. Maybe I'll put it in the feature requests.

Ric Cohn
___________________________________________________________________________

Re: Color theory and practice
Posted by: Mark Segal
Fri Jun 15, 2007 7:18 am (PST)

Jeff,

I think your general point is well-taken. I also think it important to differentiate the origin of the problem. In principle if should make a big difference whether one starts with a raw file or a JPG. As we know, a JPG made in a digital camera is pre-baked by the camera's firnware, and that pre-baking usually consists of four things photographers would normally prefer to control themselves provided circumstances permit: contrast enhancement, saturation enhancement, noise reduction and sharpening. It is quite conceivable that JPGs can emerge from the bakery with some colours saturated (either due to excessive contrast or excessive saturation boost) beyond what the printer can handle (especially matte papers with relatively low DMax), and then one is into damage control trying to compress the gamut.

Raw files on the other hand, when viewed with all the luminance, vibrancy and saturation stuff in ACR's Basic tab set to zero and the tone curves linear, can often be kind of flat through the mid-tones, but also have clipping - or if not actual clipping - "congestion" at both ends of the tone scale. I think this is in part what you mean by your second paragraph below? In these conditions, in my experience working with ACR 4, I don't need to wait for soft-proofing to do the heavy lifting - because the numerous controls in the Basic tab and the Curves tab (BOTH Point and Parametric curves being useful) really provide significant flexibility to reshape the histogram generally resulting in large improvement to the aesthetics of the image. By the time I get to Soft Proofing, it only calls for incremental adjustments to cope with the DMax compression between a satisfactory monitor image and what I would get without further ado from matte paper.

So turning to your last paragraph, I see what you mean, and on-the-whole correct, but I would slightly re-characterize it based on my recent work with ACR 4: Depending on what the image needs, more often than not I see myself both expanding and compressing the range at the same time, depending on where in the tone curve. While Dan has pointed out that an expanded number of controls does not necessarily mean better quality, in this particular case I think it does, because various combinations of luminosity adjustments between the Basic tab and the Curves tab allow one to implement gymnastics with contrast and brightness that would be very difficult to pull-off successfully with Curves alone in Photoshop, save for a lot of layer masking and blending.

Mark Segal
___________________________________________________________________________

Raw cherries file
Posted by: "Ric Cohn"
Fri Jun 15, 2007 6:48 am (PST)

Andrew Rodney graciously offered his iDisk for the raw file of 2 cherries that I mentioned a few days ago. If you are interested Andrews iDisk is: thedigitaldog. He's given more detailed directions in past posts. He gave a password of "public", but I didn't need it (Andrew do you want to change your iDisk settings?).

The cherries (for reasons I can't remember now) are only in the middle of the frame, but since it's from a 39 mp PhaseOne H45 it's still plenty large after cropping for critical viewing. What we're looking at (3 channels with drastically different ranges) is so common that anyone with their own digital captures probably could find one of their own to look at instead.

Because support for opening PhaseOne files is new, you need ACR from either CS3 or Lightroom to open it.

In any case, I first adjusted the image in Lightroom with an S-Curve and opened it, then I opened the image "flat" in Photoshop and made a master channel S-Curve in an Adjustment Layer where I tried to match the contrast look of the Lightroom image. I believe it's clear that the Photoshop Curve does lead to somewhat more color, but I see them as very similar. Changing the Adjustment Layer's mode to Luminosity results in a huge subduing of the color which gives an idea of just how much color is added by the single channel move.

In light of Jeff Schewe's excellent posts (which I haven't had time to completely study yet) it would be interesting to know whether the hue and saturation adjustments he suggests would prevent the damage to the file that Dan has pointed out. If so this definitely suggests I should be using the negative numbers more frequently. In light of the extra time needed and the lack of precision in ACR (at this point in time), I still don't think that this warrants changing my current workflow of treating ACR (and all other Raw processors) in a similar way to how I used to acquire scans, which is to say with major corrections made but with lots of "headroom" and then making these kinds of adjustments in Photoshop.

Thanks again Andrew!

Ric Cohn
___________________________________________________________________________

Re: Raw cherries file
Posted by: "Stephen Marsh"
Fri Jun 15, 2007 1:07 pm (PST)

Because support for opening PhaseOne files is new, you need ACR from
either CS3 or Lightroom to open it.

Or a hack workaround for folk like me who use older versions at home - the latest DNG converter is all that is needed, then one does not need the latest version of ACR or APL to work with the DNG.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: Color theory and practice Now sharpening in Camera Raw
Posted by: "John Arnold"
Fri Jun 15, 2007 7:19 am (PST)

Jeff,

Thanks very much. As everyone becomes familiar with the new controls, it will be interesting to hear how they incorporate them into their workflow.

John Arnold
___________________________________________________________________________

Re: Color theory and practice
Posted by: Dan Margulis
Sat Jun 16, 2007 11:29 am (PST)

Jeff Schewe writes,

Well, I see your Luminance Curve and raise you one Camera Raw adjustment
that comes REAL CLOSE to your Luminance only curve done in Lab...
I've taken the liberty (with Dan's permission to access and use the original
sRGB jpg image) my whack at producing a contrast adjustment that pretty much
matches your Luminance curve adjustment, done all inside of Camera Raw 4.1.

This version just trades in one problem for two others. The curve, as we know, forces a lot of saturation into the lighter areas of the cherries, but not to the darker areas. Your new version compensates by turning down saturation, but ACR's control of this doesn't discriminate by darkness. Consequently, instead of a file with good dark reds and oversaturated light ones, it has produced good light reds and undersaturated dark ones.

The main complaint about the first image was that the left half of the top left cherry had lost detail and had become bright red. We still see the remnants of that in the revised version but I have to agree that it's close enough, although I wouldn't accept anything redder.

The price for getting the color close here was paid in the darker colors. In the center of the middle cherry, which is where the viewer's eye naturally falls, the LAB equivalent of the sRGB value in my luminosity version is 36L52a42b. In your version the same point measures 36L48a38b. Considering that we both agreed that the color of the luminosity version was a bit tepid, we shouldn't be graying it any further.

And, as the color deepens, the problem gets worse. In the left side of the same cherry, the luminosity version reads 25L43a34b (119r15g8b); yours is 26L37a28b (113r28g20b), not even close.

The other problem: the dark pink reflection on the left of the top left cherry is now close enough, as noted. However, this and six other cherries show reflections that are much lighter. In trying to regain some of the contrast in the darker one, and in decreasing saturation, these multiple light areas got hammered--they're no blown out with virtually no color. These areas aren't great in the luminosity version but they would surely get the job rejected in the new version.

For these reasons, I'd say that this new version confirms what I said about the old one: if the objective happens to be to match the luminosity-curve version, the mandatory saturation boost in ACR makes it impossible to do so to a reasonable quality level.

I think we're in agreement on the basics here. Creating a full range in ACR saturates lighter areas more than it does dark ones. This can lead to a loss of detail in those areas. There is no way for the operator to compensate in many cases. This is as stated in PP5E and is contrary to what is stated in other books. Whether this is called a master curve or a jalapeño pepper is not a major area of concern to most, I imagine.

So much for the technical stuff. The remainder of the post made some interesting philosophical points, so I will respond to that separately.

Dan Margulis
___________________________________________________________________________

Re: Color theory and practice
Posted by: Jeff Schewe
Sat Jun 16, 2007 1:27 pm (PST)

Dan Margulis wrote:

This version just trades in one problem for two others. The curve, as we
know, forces a lot of saturation into the lighter areas of the cherries, but not
to the darker areas. Your new version compensates by turning down saturation,
but ACR's control of this doesn't discriminate by darkness. Consequently,
instead of a file with good dark reds and oversaturated light ones, it has produced
good light reds and undersaturated dark ones.

Good point...I was able to bring back some additional deep red saturation using Split Tone to punch deep reds and balance only to the darker regions...I didn't think of that in the first go round...(kinda reminds me of working with an art director...I make a change, the AD responds and I gotta make another change-if Dan stays true to form, he'll have me back at the begining where we started :~)

Regards,
Jeff Schewe
___________________________________________________________________________

What Is the Photo? (Was: Color theory and practice)
Posted by: Dan Margulis
Sun Jun 17, 2007 1:03 pm (PST)

This responds to the second half of Jeff Schewe's recent post. Since its first half was about a specific image and has little to do with what comes next, I responded to that earlier in a separate post.

And this is indeed a pretty important topic. And, I've tried to reconcile
Dan's contention that Camera Raw is not a professional level tool with my
perception that's it's a really great & valuable tool for digital
photographers...how could our perceptions be so far apart?

In principle this is a good question that can lead to useful discussion, but it's too big a brush. First, my objection to ACR is specific to the way that it establishes range--I'm not trashing anything else about the module. What I say is that the correct allocation of range is of such surpassing importance to image quality that if it is implemented improperly most of the rest of the module becomes of questionable utility. By this, I mean that if it turns out that, say, the new Vibrance setting is poorly implemented, it really wouldn't affect anything else, we just wouldn't use that particular setting. But if range is set incorrectly, it affects *everything*.

Second, it should be reiterated that, while the inadequate range setting has not been corrected in the current version of ACR, my book PP5E studied an earlier version. It stated, "Raw modules are currently the hottest area of imaging-software development. It should be anticipated that capabilities and speed will improve rapidly over the coming years." That certainly occurred with the latest release.

Then a realized something...Dan isn't a photographer. His impressive career
and experience and expertise is based primarily on taking film-something
that already fixed the original scene contrast-and correcting problems
(often caused by poor photographic technique).

Yes and no. It's true that a prepress person tends to look at an image with a different eye than a photographer does. One often concentrates on finding defects, the other looks at the whole effect. It is also true that the film--if it was in positive form, that is, set the *color* of the image, more or less.

It is not true that the *contrast* of the image was set--in fact, the contrast found in the original film was of little importance. The operator would force the scan to cover nearly 100% of the lightness/darkness range in the output file, regardless of the percentage of lightness/darkness range found in the film. Also, the operator basically ignored the internal contrast of the image in favor of emphasizing certain details.

My whole photographic career has been fighting the battle of getting the
scene-often far wider than can be captured-first on film and then later on
digital capture. So, what Dan doesn't really understand is that in the grand
scheme of things, the battle of capturing the original scene is far more
likely to involve _REDUCING_ the scene contrast to fit into the more limited
range of film or sensor, not increase it-particularly when you factor in
final output which has an even more reduced dynamic range.

On the contrary--there is nobody in the world more closely associated with this concept than I am. It seems so obvious today that we have to make intelligent decisions about what to emphasize and what to sacrifice as we move into a smaller output realm that it is difficult to believe that less than 15 years ago people did not realize it. The prevailing sentiment in the early 1990s was that if the photographer's original didn't show a full range, then the output file shouldn't either, and that the remaining contrast should be allocated in exactly the same proportion as in the original film. This was part of the philosophy I termed "calibrationism."

My August 1994 column, "Color, Curves, and Horsetrading" was AFAIK the first public statement of the idea that one should identify interest objects and exaggerate their detail at the expense of areas of lesser importance. Today, it's standard practice, as Jeff indicates above. Everything that I've written since has continued to focus on knowing what happens when we go from large to small and how to cope with it.

And, as a direct result of that main problem-too much scene contrast-Camera
Raw has been designed, first and foremost, to excel in range reduction, not
increase...hence the attention to things such as Fill Light and Recovery.

Fill Light, Recovery, Shadow/Highlight, and the blurred overlay blends that I show in PP5E are all 21st-century. They combine range extension with a type of sharpening. Nobody could do these kinds of things with highlight and shadow detail AFAIK before they made their appearance.

Important as enhancing highlight and/or shadow detail is, however, it's dwarfed by the need to enhance detail in interest objects that lie somewhere in the midrange. The traditional way of doing this is by channel curves, which ACR lacks. There's no reason to *limit* oneself to channel curves in 2007--as the commands above illustrate, we can design technically better ways. But ACR has nothing except this weak Curve command. making it, IMHO, noncompetitive overall in range enhancement even with 1980s scanners, despite the fact that it enhances highlights and shadows better.

Those times when I, as a photographer, encounter really, really flat light
and actually NEED to increase the range (contrast) a slight boost in
saturation is not only not bad but desirable.

True. The Shadow/Highlight command works this way by default, for example. But in Shadow/Highlight, the user can choose to disallow the saturation boost. One would think that ACR, which is in principle much more powerful and precise, would do the same. But it doesn't.

Which is why Thomas says he
decided NOT to use a luminance only curve but one that would more closely
resemble film being increased in contrast. He did it on purpose.

The issue is not whether this should be the default behavior, The issue is that, unlike any professional-level program, ACR does not offer a way to get around it when necessary.

Jeff has observed that the ACR curve resembles the response of, say, Kodachrome. Fair enough. But Kodachrome's behavior was not ideal for every situation. Photographers knew to use it for some shoots and not for others, because its strengths in one setting might be weaknesses in another.

The ACR method is equivalent to telling the photographer, you must use Kodachrome *all* the time. Wedding, funeral, studio, people, landscapes, outdoors, sunlight, fluorescent, light, dark, it doesn't matter. This is the film that's best for most images, therefore you must use it for all images. Works great--as long as the type of images you shoot are the type for which this kind of response is appropriate.

Dan's point of view, coming from where he is coming from is understandable
(even if he's not a photographer)...he's comparing Camera Raw to tools he's
used and is familiar with...and so am I.

I think this is entirely accurate, to the point that Jeff and I probably don't agree even on what constitutes the original image.

There are three likely variants for this image. Variant One is as opened from ACR or other raw module with as little manipulation as practicable. In ACR, I suggested that it was best achieved by zeroing out everything. It typically looks flat and may have color issues as well. Variant Two is a better-looking image created by using ACR's functionality. Variant Three is the final product, after manipulation in Photoshop.

As pre-press, I am completely focused on the final product, Variant Three. I do not give a hoot whether Variant Two looks better than Variant One. My only concern is which one will produce a better Variant Three. I am willing to create Variant Two if it helps that goal, and to forget about it otherwise. To me, Variant One is the original. Extending its range is necessary, because it's too flat.

As photographer, it's doubtful that Jeff even acknowledges the existence of Variant One. He is willing to proceed directly to Variant Two, because that's in keeping with his experience. The image is exposed on whatever film happens to be in the camera. If that's the wrong type of film for the job, such is life: one couldn't load the camera with six different brands of film and press a button to choose on which to expose the next shot.

Also, photographers understandably wish to have their original product look good, as it usually did in film, even if they know that there will be further alterations down the line. Variant One looks bad, so naturally it is foreign to a photographer to tell the world that *this* is his original--he much prefers the look of Variant Two. The traditional product of a photographer is *not* a final digital file but rather a good-looking intermediate version that somebody else finalizes. A photographer, therefore, is traditionally not as concerned with Variant Three as a prepress person, and far more concerned about representing his work in a favorable light with Variant Two, as opposed to the toothless Variant One.

And so, I think that Jeff feels that the photograph does not exist until it has been interpreted to some extent, so that his original is Variant Two. That's why he's described ACR as a range *compression* tool. But me, I think that the flat-looking Variant One is the original, which is why I describe it as a range *expansion* tool.

From the perspective of graphic arts and prepress production, maybe Camera Raw isn't
the tool for you. Course, we are fortunate that we then can rely upon
Photoshop to do those kind of corrections-and isn't it convenient that when
you open an image from Camera Raw, it opens into Photoshop?

All true. From the standpoint of production, however, we are now sometimes handed raw files and asked to proceed. We *have* to start in raw, and we have to end in Photoshop. In between, during the period that we're in raw, whether we are conservative, or whether we try to make major corrections, and if so do we just eyeball it or do we try for precision--these are the questions. In Photoshop CS2, I felt that the answer was clearly to be conservative in view of the range-handling issues. With the improvements in the current version this may no longer be the case, and if it is not, I would welcome the development.

And for photographers (at least THIS photographer) Camera
Raw is a nifty professional level raw image adjustment tool...but, it's
ain't Photoshop.

Not for a while, anyway. I appreciate Jeff's thought-provoking and instructive comments.

Dan Margulis
___________________________________________________________________________

Q: ACR/LR Split Toning uses?
Posted by: howienordstrom
Mon Jun 18, 2007 2:05 am (PST)

Jeff Schewe's post about using split toning to offset saturation issues caused by ACR & LR tonal curve adjustments (http: //tech.groups.yahoo.com/group/colortheory/message/18040) prompts me to ask a divergent question about the use of split toning.

For some time now I've been playing with LR split toning not as a method for adding an artistic color cast but rather as a method for _offsetting_ color casts in the shadows or highlights. I've not read anything (here in ACT or in other locations) that has *directly* discussed this option: if it has been addressed kindly point me in that direction.

It seems that there is a benefit to using split toning in this way--partial, or in some cases complete, removal of color casts. Certainly, as has been discussed in the thread "Color theory and practice", a more conservative approach in LR/ACR with subsequent adjustments in PS may be the most optimal (defined by individual time requirements, developer's skills, & desired results) approach to completely remove color casts, but what *are* the actual detriments to split toning for cast _removal_ with the LR/ACR engine?

As in the discussion about tonal curves in the afore mentioned thread, perhaps a discussion about split toning will shed some light on the way the tool was designed and how it can be utilized/optimized.

/Howie
howienordstrom.com
___________________________________________________________________________

Re: What Is the Photo? (Was: Color theory and practice)
Posted by: "Ric Cohn"
Mon Jun 18, 2007 6:36 am (PST)

On Jun 17, 2007, at 3:46 PM, Dan Margulis wrote:

From the standpoint of production, however, we are now sometimes
handed raw files and asked to proceed. We *have* to start in raw,
and we have to end in Photoshop.

I've had a firsthand look inside some large agency digital departments and I think the biggest potential problem is receiving files where the photographer has made what he/she thinks are improving adjustments to the raw file only to make the rendered file much more difficult to correct. Many clients have noticed that digital files frequently need much more work afterwards than used to be the norm with film. I personally hate if a client tries to insist on receiving the Raw files (in addition to my processed files) and I try to convince them that *I* know what I'm doing and the Raw files would not be helpful to them (and I hope I'm right <g>).

I'm afraid that "experts" pushing photographers to make adjustments without adequate explanation of pitfalls, and understanding of what a proper file is, and without admitting the limitations of the current tools will only exacerbate the problem and will probably leave residual mistrust that will take years to overcome. In the early days of digital I had clients that had been burned so badly by bad digital that they had sworn they would never accept it again. I had to send over DVD's of digital images to prove that it was the photographer and not the technology that was the problem. That time has mostly past, but unfortunately is being replaced by a clientele that doesn't trust the photographer to supply a professional level product.

Bottom line (for me at least): Raw processors are amazing tools that when used conservatively can greatly improve the output from digital cameras. As these products are improved, for top quality output, it will make sense to make more and more adjustments to the Raw file before deciding what additional work is needed in Photoshop.

Ric Cohn
___________________________________________________________________________

Re: What Is the Photo? (Was: Color theory and practice)
Posted by: "Andrew S. Webb"
Mon Jun 18, 2007 10:29 am (PST)

I have to correct a lot of digital files that the photographer, knowing a little about Photoshop, has rendered hideous and unworkable. Photographers are suddenly their own film labs, and many don't know anything about post-capture processing. I very often request the RAW files, and when there are protestations, I ask that they send along some examples of their work—portfolio jpegs, for example. That way I can match their personal style in post and everyone ends up happy. Sometimes I collaborate with the photographer, showing them my work before delivering to the client so that they remain in control of their own "look".

Some photographers have been very resistant to providing RAW files, and I understand this: It's like giving the client your _negatives_ instead of your prints. Others bristle at the suggestion that they aren't the absolute best post-processors of their own files; when the film lab analogy is drawn, they usually relax. I've found that many of them are well aware of their deficiencies, but are worried that they will appear incompetent in the clients' view. They tend to be grateful and pleased that there are people like me who will work with them to deliver an optimized file (sans scorn and condescension). I now have a few photographers as permanent clients who explain to *their* clients that the files will be post-processed by me and delivered by me. If you're a freelance retouching artist/prepress jockey, I suggest you talk to some of the photographers in your are and see if they're looking for this kind of assistance. It's lucrative and quite satisfying work.

Cheers,

_andrew webb
___________________________________________________________________________

Re: Color theory and practice
Posted by: "Rick Gordon"
Mon Jun 18, 2007 3:14 pm (PST)

Aren't LightRoom's and ACR's raw operations built from the same underlying code? Do the implementations vary between one and the other? I don't use LightRoom, but that's what I've heard a number of times (barring the fact that until the update appears, LightRoom is operating with the ACR 4.0 model rather than the 4.1).

Rick Gordon
--

___________________________________________________

RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________________

WWW: http://www.shelterpub.com
___________________________________________________________________________

Re: Color theory and practice
Posted by: Richard Wagner
Mon Jun 18, 2007 7:16 pm (PST)

On Jun 18, 2007, at 1:19 PM, Rick Gordon wrote:

Aren't LightRoom's and ACR's raw operations built from the same
underlying code? Do the implementations vary between one and the
other? I don't use LightRoom, but that's what I've heard a number
of times (barring the fact that until the update appears, LightRoom
is operating with the ACR 4.0 model rather than the 4.1).

What you're heard is correct, although there are differences in the GUI. For example, LR uses the Melissa color space (ProPhoto primaries with a 2.2 TRC) to display the histograms and % color for pixel values, whereas ACR uses the output profile color space for the histograms and 8-bit RGB values for pixels. The internal computational pipeline is supposedly identical now, though, with the exception that you mentioned of LR not yet having the GUI for the 4.1 changes.

This is a dilemma, since LightRoom is such a promising product,
with some great (and unique) features, and Raw Developer can't
compete in overall workflow. So for now, all the commercial work
goes through LightRoom/ACR; only select images get the Raw
Developer treatment.

I also use Raw Developer for a few difficult images. It is a great RAW processor, but I agree, for workflow, it comes nowhere close to ACR. For that matter, Nikon's Capture produced great output, but the overall program was such a slow kludge that it was not usable with any significant volume of images.

--Rich Wagner
___________________________________________________________________________

Re: What Is the Photo? (Was: Color theory and practice)
Posted by: "Paco Rosso"
Tue Jun 19, 2007 7:02 am (PST)

Some photographers have been very resistant to providing RAW files,
and I understand this: It's like giving the client your _negatives_
instead of your prints.

A raw is not a finished picture. is like the negative. But what you want to print is not the negative, but the enlarge copy. I will never give a raw because it is not the picture. The picture is the tiff. Nobdy will tell me how I have to "develop" my pictures. The final image has a contrast control, a tone reproduction. The raw has nothing. Ten differents persons working with the same raw file will make ten differents pictures. I will publish my picture, no the picture of other person.

it seems you do not understand the difference beetween a negative and a print. It is a right of the author to decide the way the picture is shown.

Paco Rosso

Luz-Color-Fotografía

http://pacorosso.blogspot.com
___________________________________________________________________________

Re: What Is the Photo? (Was: Color theory and practice)
Posted by: "Andrew S. Webb"
Tue Jun 19, 2007 8:29 am (PST)

I think that you are misunderstanding what I said. I do indeed understand the difference between the negative and the print, and that is exactly what I said in the bit that you quoted.

You said:

"I will publish my picture, not the picture of other person."

I agree that the picture you provide to the client should look the way you want it to look. The problem is that more than a few photographers who have switched to digital don't know how to post- process their files so that they DO look the way they want. They light, compose and expose the shot to make the shot in their own style, and then they make mistakes like unintentionallly blowing out hilights in Photoshop.

You said:

"I will never give a raw because it is not the picture."

That's fine if you know what you're doing. If you don't, your client will reject the picture and you will not be paid. I have seen this happen more than once.

If you go back and read my post again, I think you will realize that I am on your side and that I do not disagree with what you are saying.

Cheers,

_andrew webb
___________________________________________________________________________

Re: What Is the Photo? (Was: Color theory and practice)
Posted by: "George Machen"
Tue Jun 19, 2007 1:08 pm (PST)

"Andrew S. Webb" wrote:

... Others bristle at the suggestion that they
aren't the absolute best post-processors of their
own files; when the film lab analogy is drawn,
they usually relax. ...

Andrew, could you elaborate a little more what you've successfully said to them vis-a-vis the film lab analogy, as well as how you phrase the other aspects your post mentioned that reassures the photographers?

I currently happen to be in the middle of working up a cover letter pitching exactly what you're talking about, and while I'm not asking you to write boilerplate for me, suggestions pointing toward wording that actually has worked for you in the field would be a big help to this clumsy writer.

Thanks!

- George Machen
___________________________________________________________________________

Re: What Is the Photo? (Was: Color theory and practice)
Posted by: "Andrew S. Webb"
Mon Jun 25, 2007 6:13 pm (PST)

George,

It varies a lot. I try to do it over the phone, where it is easier to gauge reactions.

In my youth, I printed custom B&W for a lab in Chicago. I worked with the photographers a lot then, finding out all I could about the intended mood, intended use, etc. I ask the same sorts of questions now. Mentioning that you understand that the RAW file is like a negative and you will be working *with* them to produce the best possible print is important, and making them understand that you are trying to assist them in producing the clients' vision helps, too.

Below is a snippet of an email I just sent off to a photographer I know. Since we've worked together in the past (I also do on-set pre- composite consulting to make sure that all the pieces match), an email is all I need in this case. I'm not the biggest fan of email, so, as I said, I usually try to do this stuff over the phone. It's hard to sound like a real person in an email. It's hard to be convincing and reassuring. (It's the difference between XX00 and a tender kiss...)

Client asked if I could reduce shine on models' skin, and I said yes. In some cases, it might be easier for me to work from the RAW Phase1 files when I do this—I can recover a certain amount of blown- by-shine hilights during RAW processing. I didn't want to step on your toes by asking for the RAW file before I checked with you. I understand that a RAW file is essentially your negative, and that at least part of your photographic art is in rendering that file to a TIF in your own particular way. If the idea makes you uncomfortable, I would be happy to submit my retouched files for your approval before I show them to the client.

Cheers,

_andrew webb
___________________________________________________________________________

Re: Rendering the print
Posted by: dave_cardinal
Sat Jul 7, 2007 12:36 pm (PST)

I'm not sure why Dan objects to RAW except that maybe it will over time
diminish the need for his excellent techniques in repairing crap images. But I
sense that Adobe also fears RAW and promotes its own version because Nikon
Capture, which I use and paid for and receive no consideration from Nikon for
using, might some day allow me to do without Photoshop.

In my case, I certainly don't object to Raw (it is almost all I shoot) and I don't think Dan does either, but I did find the JPEG-bashing in the cited paper unfortunate, as I think that otherwise the paper is quite a nice explanation of many topics with some innovative illustrations.

In particular, to choose a poor image of the moon & show how it can be saved in Raw is one thing, but the casual reader of the explanation in the paper might be left with the impression that one can't shoot the moon using JPEG--which is not true--the image shown just happens to be poorly exposed and therefore hard to correct.

Then to compare JPEG to a print is also a little harsh. Sure it isn't as flexible as raw, but it is also not anything nearly as fixed as a print of course. Even artifacting has gotten pretty minor with the current high-end cameras.

The comment about JPEG "setting the exposure window" is also a little misleading, since both JPEG & Raw set the exposure the same (of course) and it is really the tone mapping of the values within that range that can be changed (yes, so that lets us tweak the apparent exposure in our raw processor a stop or two, but we aren't really able to go back and magically change our camera exposure after the fact).

I wish those interested in promoting Raw could be a little more even-handed. When I lead photo trips one of the big issues we have with new photographers is that some of them have read too many "scare stories" about JPEG and are certain that they have to shoot Raw Day 1, before they really understand exposure, basic camera technique or have a digital workflow established. Raw can be an awesome teaching tool, but it can also be intimidating for those not already comfortable with their camera & computer.

Ironically we then get statements like this from Canon: "The net result is that even if the G7 offered raw image capture...there would be no discernible improvement in image quality compared to...superfine JPEG mode," Westfall <Canon VPsaid--which just seems odd since presumably all the same benefits of Raw should apply even if the G7 pixel size is smaller and therefore noisier.

--David Cardinal

PS Just for a quick reference, there is a shot of the moon I took several years ago with a Powershot A80 (now obsolete) online at:
http://www.pcmag.com/article2/0,1895,1819577,00.asp
Even after re-sizing and rendering on the web it still shows more detail than the "JPEG" in the paper.

--David Cardinal
Cardinal Photo / Pro Shooters LLC
http://www.cardinalphoto.com
http://www.nikondigital.org
http://www.proshooters.com
___________________________________________________________________________

Re: Rendering the print
Posted by: Dan Margulis
Sat Jul 7, 2007 10:02 pm (PST)

John Denniston writes,

I'm not sure why Dan objects to RAW except that maybe it will over time
diminish the need for his excellent techniques in repairing crap images.

On the contrary, AFAIK I am the originator of the concept of raw correction, because I realized very early in the age of digital photography that the cameras' internal algorithms occasionally damage image files so badly that they become difficult or impossible to correct. So, I have shown in PP5E that certain images need to be opened in Camera Raw or a similar module, to prevent this damage.

I certainly do not object to transferring more complex corrections to the raw module *provided* that there is no loss in quality by doing so. At present, this is not the case. We had a long thread on this recently that there is no need to rehash. The implementation of range-setting within Camera Raw is too primitive, IMHO, for it to be an acceptable choice for a quality-oriented user.

If that changes in the future, then it's a different story, as is always the case when products improve. The current rev of Raw is much better than the CS2 version that I tested. Among the changes in my recommendations that may be seen in the future are:

1) At Photoshop World in September, I will be showing a new workflow. It is capable of spectacular results, but it puts the file under considerable stress, so much so that compression artifacts of JPEGging become visible. My testing is not complete but I think it is likely I will have to warn people to shoot raw (or uncompressed TIFF) where possible, if they are to use these new techniques.

2) In CS2, it was not possible to use Camera Raw for anything other than a raw file. In CS3, we can use it on any type of file, if we like. When we have the type of JPEGs that exhibit the color damage that raw captures can avoid--overly saturated light colors, casts that aren't uniform from highlight to midtone--then the deficiencies of Camera Raw are not relevant, because the damage has already been done. The only question is how to make the best of the bad situation. It looks to me like there is promise in trying to open such images in Camera Raw and use its command structure rather than Photoshop's, as a start in repairing the image.

Dan Margulis
___________________________________________________________________________

Re: Rendering the print
Posted by: "Chris Brown"
Sun Jul 8, 2007 8:17 am (PST)

I must be living under a rock or slab of concrete. I thought every pro photographer shot in raw mode. Given the amount of flexibility and control to achieve such a wide range of exacting results in post, I shoot raw exclusively on every job. (I shoot JPEG, though, when scouting locations, during cattle calls or when using my wife's mighty Canon PowerShot A400).

If the paper [Rendering the Print] was a propaganda piece for Adobe intending to spurn the masses to shoot in raw mode, I doubt it'll work. Just hang out at your local photo store for a while and listen to the moms and grandparents complain that they can't "just get a damn print" off the computer. The last thing the masses want to do is spend hours in front of their computer neck-deep in parametric editing.

Cheers ~
Chris Brown
___________________________________________________________________________

Re: Rendering the print
Posted by: "dmargulisnj"
Tue Jul 10, 2007 8:04 am (PST)

Chris Brown writes,

I must be living under a rock or slab of concrete. I thought every
pro photographer shot in raw mode. Given the amount of flexibility
and control to achieve such a wide range of exacting results in post,
I shoot raw exclusively on every job. (I shoot JPEG, though, when
scouting locations, during cattle calls or when using my wife's
mighty Canon PowerShot A400).

In my classes I get an ever-increasing number of ultra-serious hobbyists, typically successful professionals/businesspeople, 50+, who happen to be zealots for good images. Surprisingly, these people are more likely, in my experience, to be using raw than professional photographers are--close to 100%.

Those photographers who haven't adopted raw often are shooting large volumes of material and/or have extremely tight deadlines. Others haven't been willing to make the effort to learn and are satisfied without it.

But I think the most interesting subcurrent is, I hear anecdotes about photographers who have a real issue with putting the raw file in somebody else's hands. One photographer told me, for example, although he himself shoots raw he denies it to his clients, so that they will not ask him for the raw files. Also, photographers who give out several raws from the same shoot report that their clients unerringly pick out as the favorite the one image that has a non-obvious technical defect that makes it nearly impossible to fix.

If the paper [Rendering the Print] was a propaganda piece for Adobe
intending to spurn the masses to shoot in raw mode, I doubt it'll
work. Just hang out at your local photo store for a while and listen
to the moms and grandparents complain that they can't "just get a
damn print" off the computer. The last thing the masses want to do is
spend hours in front of their computer neck-deep in parametric editing.

That's right. In any event "the masses" are unlikely to be able to exploit the benefits of shooting raw, at least as far as image quality is concerned. OTOH, as the raw modules become more user-friendly and offer benefits in organizing images, they'll increase their appeal. I suspect that progress depends more on the camera vendors than it does on Adobe. Those vendors have their own ideas on how things should play out, and steering people toward manipulation in Adobe or Apple products is not real high on their priority list. The next few years should be quite interesting.

Dan Margulis
___________________________________________________________________________

Re: Rendering the print
Posted by: "Lee Clawson"
Tue Jul 10, 2007 3:50 pm (PST)

Henry,

I'm confused. Why would anyone think of giving RAW files to any print shop. That's like giving an exposed roll of film still in the canister and un-processed to a scanner operator.

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: Rendering the print
Posted by: "Bob Smith"
Wed Jul 11, 2007 4:25 am (PST)

On Jul 10, 2007, at 9:22 PM, Lee Clawson wrote:

I'm confused. Why would anyone think of giving RAW files to any
print shop. That's like giving an exposed roll of film still in the canister and
un-processed to a scanner operator.

I wouldn't think of it, but on rare occasions a client has and has asked me to deliver raw files. In ten years of shooting raw, I could probably count on my fingers the number of times this has happened. All of these instances have been within the last two or three years. I've always refused. Just as I would never turn over a neg. I'd have zero control over what happens to it, but I'm sure I'd be the target of blame if the end result is sub-par. As programs like Lightroom or Aperture become more routine I'm sure photographers will see more client requests for raw format files.

Bob Smith

Accurate Image • Bob Smith Photographer • Waco Texas USA
http://www.accurateimage.org
___________________________________________________________________________

Re: Rendering the print
Posted by: "Alcides Pomina"
Wed Jul 11, 2007 6:25 am (PST)

 The reason for the request of RAW files by some publishers is to verify that the content of the final image has not been altered. As you may remember some photojournalist enhanced the drama of some war shots taken in Iraq.

Al Pomina
___________________________________________________________________________

Re: Rendering the print
Posted by: "RJay Hansen"
Wed Jul 11, 2007 8:19 am (PST)

One should never be surprised by what may be submitted to a print shop as artwork; I wouldn't rule out anything.

RJay
___________________________________________________________________________

Re: Rendering the print
Posted by: "Henry"
Wed Jul 11, 2007 10:08 am (PST)

A pro probably wouldn't hand off raw files unless they wanted and agreed to allow creative editing by the party that received it. In this case, there is hopefully a clear understanding of the situation.

There are a number of reasons, not good ones, that raw files are exchanged. Most reasoning is due to misunderstanding - but some reasons go a little deeper. I have personally been told by creatives that a professional, such as myself, "should be able to handle" raw files. Imagine that!

This mostly boils down to a dodge of responsibility for the printed outcome of an image. The reasoning runs like this:

print buyer/creative/photographer: I have delivered the very best information, state-of-the-art. It is your responsibility to render printing that matches my expectation. Their expectation *might* be the jpg or the preview that they saw, but that would just be a guess.

We won't even go into cropping, sizing and other tasks that are avoided when raw files are handed off.

Lee, you understand the nature of raw - that they are not ready to print. But in the world, there is a degree of "honest" misunderstanding about raw. Some people earnestly believe that they are doing the right and best thing by handing off raw files. Some people actually shoot raw+jpg but haven't the software or the skills to process the raw files. They submit raw as a matter of course, innocently believing that this is what is expected. Their numbers are growing as more cameras support raw.

A similar progression occurred with the growth of RGB as an exchange color space for offset print work. At one time it was rare that a print shop would receive an RGB image.

Henry Davis
___________________________________________________________________________

Re: Rendering the print
Posted by: "Jim Bean"
Wed Jul 11, 2007 11:11 am (PST)

certainly if there were to be a problem, the person on the bottom of the food chain (many times the photographer) catches the heat.. but here is a real trend. the clients ask for the raw file.. why? they attend worshops or read trade publications where their peers/experts demonstrate/discuss raw capabilities and it is assumed that the better shops always request/demand the raw files..therefore it becomes an 'industry standard' to make this request..

just as the ealier post regarding a client choosing a 'less than perfect' original raw... it is important to never give clients choices that you do not want them to make. I personally have never delivered a raw but I would if 'appropriate'...

jim bean professional photographer
___________________________________________________________________________

Rendering the print
Posted by: "Richard Chang"
Wed Jul 11, 2007 1:42 pm (PST)

Why would anyone want to give a RAW file?

If one is comfortable with the vision and the craft of the image, why not?

If one is to believe the Adobe Camera RAW and Lightroom marketing, anyone can get a great result with a RAW file. The internet is rife with opinions and techniques guaranteed to make a perfectly good looking picture-- all you need is a properly profiled monitor and the appropriate output profile to work with; I've read it here, posted by the color management cognoscenti. And, with no brainer color management procedures in place, why shouldn't they make just as good a picture as you do?

If color management has it's way and has universal adoption of the policies that make great pictures, why are you better equipped to make a great picture? Is it just because you ritually filled the histogram?

It would seem that the marketing of high tech image editing on this list is less about craft and more about technology, which has a habit of changing-- what was required a few years ago is passe today. It also seems to me that that where some folks think there's more than one way to do things, there are a few contributors who feel that they're right and anyone with a different approach is wrong-- I know that I've been told that I'm wrong with regard to exposing pictures.

Richard Chang
___________________________________________________________________________

Re: Rendering the print
Posted by: "Henry"
Wed Jul 11, 2007 1:50 pm (PST)

I'll suggest that in the case of raw images, the photographer is assumed to have delivered the very best, state-of-the-art, image information. If this is how raw is being misunderstood, the photographer probably won't be blamed for the outcome. This is not the same, but is akin to the photographer's innocence when delivering RGB in certain situations. The photographers client has seen beautiful RGB renderings, so the fault presumably belongs whomever has mishandled the file.

Your observation that clients and other parties ask for raw files is another testimony to the misunderstanding that I mentioned. I would have hoped that the "Rendering the Print" paper and others like it would make a more prominent explanation and warning that raw files are not ready to print. To me, this should be the very first thing discussed in such writings.

Understandably, a product endorsement wouldn't want to kick off with warnings about what the product doesn't do, or how dangerous it might be. All of the good claims about Camera Raw are worthy - it is a very good format and so forth. Camera Raw workshops, seminars and papers could do more to educate users about the more obvious aspects of the format.

Henry Davis
___________________________________________________________________________

Re: Rendering the print
Posted by: "Lee Clawson"
Wed Jul 11, 2007 5:14 pm (PST)

on 7/11/07 12:55 PM, Henry wrote:

But in the world, there is a degree of "honest"
misunderstanding about raw. Some people earnestly believe that they
are doing the right and best thing by handing off raw files. Some
people actually shoot raw+jpg but haven't the software or the skills
to process the raw files. They submit raw as a matter of course,
innocently believing that this is what is expected. Their numbers
are growing as more cameras support raw.

A similar progression occurred with the growth of RGB as an exchange
color space for offset print work. At one time it was rare that a
print shop would receive an RGB image.

Henry,

The delivery of RGB was at least something I could understand. This really surprises me. I really had no idea.

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: Rendering the print
Posted by: "Ron Kelly"
Wed Jul 11, 2007 5:17 pm (PST)

Richard:

I think you make a good point, if I understand you correctly.

We've heard many times that RAW is the best, better than TIF and certainly tons better than JPEG, only someone living under a rock would shoot JPEG, etc.

I would agree with those who don't want to hand off a RAW file. So what do you supply? A TIF, of course, except now the photo buyers and art directors have heard of RAW and they've been told "it's the best" so they don't want to settle for some other crummy format.

Once a RAW file gets to an art director . . . maybe he or she will know better than me how to use it. But I highly doubt it. I've taken classes from Dan Margulis and worked years to develop my skills. Based on my limited experience, most art directors/photo buyers are not highly skilled in this area. They don't know that, however, and that is the biggest problem.

Are you suggesting that color management will steer them safely through the RAW stage? I think you were being sarcastic, but I'm not really sure. (I don't have the correct profile to sniff out sarcasm all the time <g>)

The value of my photography is highest when the final result looks it's best. If the printed piece doesn't look good, I'm ultimately going to pay the price no matter who messed up along the way.

It would seem that this business of handing off RAW files, if it becomes the norm, will be a disaster.

Ron Kelly
___________________________________________________________________________

Rendering the print
Posted by: "Richard Chang"
Wed Jul 11, 2007 8:58 pm (PST)

Ron:

We've heard many times that RAW is the best, better than TIF and
certainly tons better than JPEG, only someone living under a rock
would shoot JPEG, etc.

I think that Jpg is no worse than film-- it is a singular development which is problematic if that development is inconsistent with the rendering intended, or the lighting of the original scene. If the scene illumination and the rendering target are in line, then a reasonable result should be at hand. We shot film within limits and there's no reason why we shouldn't be able to shoot jpgs under controlled conditions.

Are you suggesting that color management will steer them safely through the RAW stage? I think you were being sarcastic, but I'm not really sure. (I don't have the correct
profile to sniff out sarcasm all the time <g>)

It was supposed to be tongue-in-cheek but I guess I botched it. I apologize to anyone who was offended; I'll keep my warped humor to myself in future posts.

What I was trying to suggest is that if everyone followed the approved color management methodolgy, then there shouldn't be a reason why you couldn't give off a RAW file. Experts on this list have asked for RAW files to prove that they know how to do it right, that their approach is the prudent workflow. If everyone knows this approved approach, there should be no reason why a RAW file couldn't be safely handed off, right?

It would seem that this business of handing off RAW files, if it becomes
the norm, will be a disaster.

That's an interesting observation. I'd bet there's a quite a few readers on this list who believe they know how to handle a RAW file. And I'll bet there are folks who know a lot more about Camera RAW than I do. Does that superior knowledge guarantee that they can make better images than I can, with my files?

What if you did an image edit and saved a file with an embedded profile? Could you supply your imaging intention with the RAW file and trust that your vision (like a good looking chrome) would be safe? Would you give a set like this (I would) to Dan? How about to an unknown stranger? Could giving a RAW file work? That may depend on who you're giving it to.

Richard Chang
___________________________________________________________________________

Re: Rendering the print
Posted by: "Richard Kenward"
Thu Jul 12, 2007 7:43 am (PST)

In his posting of Wed, 11 Jul 2007, Richard Chang writes

I think that Jpg is no worse than film--

Dear Richard

I will confine my reply to the above. I do not understand your statement, perhaps you could expand on it please. Film has all the detail available in a pure form and does not suffer from compression degradation artefacts. True it has grain...perhaps that's what you mean.

Cheers

Richard Kenward
___________________________________________________________________________

Re: Rendering the print
Posted by: Mark Segal
Thu Jul 12, 2007 7:43 am (PST)

Richard,

My late father had a wonderful expression to the effect that "you never show a fool a job half-done". A raw file is a job half-done.

Capture is the first stage, raw processing the second stage and output processing the third stage in the creative process. Colour management is purely a technology designed to achieve consistent rendition of colours between devices that would produce different colours for the same file values absent the device characterization that the profile provides. But the photographer/artist needs to know what those values should be. That is vision and judgment - a separate role and function from colour management.

The answer to your question below about "guarantees" is NO. But it is guaranteed that if you have all the raw data and you know what you are doing with your raw processor you should be able to produce superior images under a whole variety of conditions in which your JPEGs would pale by comparison.

Mark Segal
___________________________________________________________________________

Rendering the print with raw files
Posted by: "Richard Chang"
Thu Jul 12, 2007 12:13 pm (PST)

Mark:

My late father had a wonderful expression to the effect that "you
never show a fool a job half-done". A raw file is a job half-done.

A raw file on its own, is indeed a job half done. That's why I recommended the shooter process an rgb with an embedded profile so that the downstream workflow would know what the shooter wanted the image to look like. I don't think that anyone would contest the idea that a raw file is the best place to start. The shooter knows what he wants, but is he the best person to produce good looking press output?

Not all photographers are adept at preparing a file for CMYK output onto an unknown press. I will assume there are readers here who have not taken Dan's Color Theory I and II. I will also assume that the complexities of Andrew Rodney and Chris Murphy's explanations of the pitfalls in color management have not been completely understood by all (myself included). In the old days we gave a chrome that we were comfortable with, without too much complaint.

Capture is the first stage, raw processing the second stage and
output processing the third stage in the creative process. Colour
management is purely a technology designed to achieve consistent
rendition of colours between devices that would produce different
colours for the same file values absent the device characterization
that the profile provides.

I agree.

But the photographer/artist needs to know what those values should be.

I believe this too, but I'm troubled by the number of times where I've read "we don't know where the job will print, with what press, and on what paper"!

That is vision and judgment - a separate role and function from
colour management.

I think that successful vision and judgement require an understanding of the target. I am only suggesting that as color managment becomes more mainstream, the gross errors that could occur, should not occur. While I am not in the habit of giving out raw files myself, I have to ask both why? and why not? The principle problem I have with giving a raw file is the downstream people don't know what it should look like-- Camera Raw's default is not necessarily what I want. But if we believe in embedded profiles (this list is rife with reasons why we should), why can't we include the look of what we want with our raw file by including an rgb that we're satisfied with?

The answer to your question below about "guarantees" is NO. But it
is guaranteed that if you have all the raw data and you know what
you are doing with your raw processor you should be able to produce
superior images under a whole variety of conditions in which your
JPEGs would pale by comparison.

I am not under the impression that raw is somehow inferior-- at MegaVision we implemented a raw workflow in 1996, with separate chrominance and luminance adjusts. And I believe that under variable conditions, a raw file is indeed more flexible.

What I did suggest, is that for a singular, known condition, a jpg could be made useable with lighting and post production for that condition. Perhaps I spoke out without thinking it through, but I still believe that if I can light for a film emulsion, I could construct a lighting scenario for jpg. Would it be flexible? Not likely, but it should be as good as a decent chrome emulsion. Could I light a jpg for one target that I know about? I'll bet I can.

Richard Chang
___________________________________________________________________________

Re: Rendering the print
Posted by: "Andrew S. Webb"
Thu Jul 12, 2007 8:47 am (PST)

I think that handing off RAW files to clients is bananas. Handing them off to retouchers is a different thing, though I would make sure that I had a good relationship with the retoucher and/or had final approval of the resulting files.

Clients (ADs and CDs and even many Production Artists at agencies) often don't have a grip on how to handle digital images. I've seen them screw up files even when there was an explanatory note attached.

At least retouchers are trained (theoretically) to handle RAW and to understand that it most often isn't as straightforward as convert-and-print.

This discussion has got me worried. I've been thinking of writing a simple Best Practices paper on digital image handling and file types (too many "designers" editing and re-saving JPEGs over and over), and maybe I'll add an explanation of why, as a client, you shouldn't ask for RAW files.

As far as Richard's question goes, no, I don't think so, _unless you were sitting next to me talking to me about what you wanted_. A better-trained person might indeed produce a better result given the correct direction.

I think the point is more that there are many photographers who have very little digital knowledge, and they're left adrift with a RAW file they don't know much about. In that case, the "superior knowledge" will beat out the shooter every time. This is why more and more commercial photographers are hiring "digital techs", who help them to realize their vision in the new medium.

If you're knowledgeable and talented and want a job in this new field, you're in a perfect position today. Go talk to local photogs and show them what you can do. Find a flexible RAW file in their style and render it three distinctly different ways. Point out that they never had to print their own color before, so why should they now?

Cheers,

_andrew webb
___________________________________________________________________________

Re: Rendering the print
Posted by: "Dennis Dunbar"
Thu Jul 12, 2007 2:07 pm (PST)

There are certain times when delivering the RAW file does work better than a processed one. When the photographer does not know what the final use of the image will be, or especially if the image is an "element" in a more complex composite.

For instance in the work I do retouching/finishing art for movie posters it becomes very clear that the photographer could not imagine how the image will be used in the final design. The images used are really elements that are combined. tweaked and twisted into something completely different from what they started out as. Often the images come from photography done on the set while filming is taking place. How can the shooter know the head from the outdoor shoot on the river will be used on a body double and set into a night scene?

While I admit these uses are the exception and not the norm I do want to say there are certainly times when it is legitimate, even preferable to provide RAW files. As a retoucher/finisher I ALWAYS prefer RAW files to processed ones. They just give me more options to get to where I need to with the images.

Dennis Dunbar ;-)>
___________________________________________________________________________

Re: Rendering the print
Posted by: "Nick Dunmur"
Thu Jul 12, 2007 2:19 pm (PST)

On 12/7/07 16:33, "Andrew S. Webb" wrote:

I think that handing off RAW files to clients is bananas.

I agree entirely with that, but...

Point out that they [photographers] never had to print their own color before, so why should they now?

Disagree entirely with that...as far as I©ˆm concerned, learning as much as I can about digital capture, process and colour management enables me to charge for these additional skills. I can produce images that I think are Œright©ˆ for the job, add some post-production, an RGB to CMYK conversion if required and finally a proper FOGRA certified proof. It adds value to what I do and makes me more indispensable to my clients plus enables me to make more money from any given job. In an area of industry (photography) that is continually being devalued by microstock, Royalty Free and the (misplaced) notion that because it©ˆs digital, it must be (a) cheaper and (b) easier (so the client can have a go), it makes total sense to me to have as much control as possible, as long as it©ˆs properly costed and charged for. This is where a lot of people come unstuck....not charging for all the additional work required to get from digital capture to press-ready file.

Kind regards

Nick D
--
Nick Dunmur - Photographer
Nottingham, UK
+441159509685
___________________________________________________________________________

Re: Rendering the print
Posted by: "Ron Kelly"
Thu Jul 12, 2007 4:47 pm (PST)

Dennis:

That would be a fun business to be in, and yes, I certainly see that RAW is a better starting point for you. It's certainly a long way from the type of work that I do. I'm into pretty straight stuff like believable colours, clean neutrals, etc. and I like to think that I alone control how the image is manipulated, not someone down the line from me. But that is my business model, admittedly.

Once again we have a discussion thread that proves that what works for one is not desirable for all. It's only natural to assume that everybody else wants to do what you do, too.

However, don't you feel that putting someone's head on another body and spinning the clock twelve hours or so is pushing things a little far? You mean to tell me that some of these posters are "enhanced?" I never would have believed it.

Cheers,
Ron Kelly
___________________________________________________________________________

Re: Rendering the print
Posted by: "Werner Tschan"
Fri Jul 13, 2007 3:22 am (PST)

I totally agree with the below.

1) The money ends up in my pocket.
2) Nobody is going to interpret my pictures other than I intended.
3) The responsibility is clearly with me.

Werner Tschan - Photographer, Switzerland

Nick Dunmur schrieb:

...as far as I©ˆm concerned, learning as much as I
can about digital capture, process and colour management enables me to
charge for these additional skills. I can produce images that I think are
OEright©ˆ for the job, add some post-production, an RGB to CMYK
conversion if required and finally a proper FOGRA certified proof. It adds value to
what I do and makes me more indispensable to my clients plus enables me to make
more money from any given job. In an area of industry (photography) that is
continually being devalued by microstock, Royalty Free and the (misplaced)
notion that because it©ˆs digital, it must be (a) cheaper and (b) easier (so
the client can have a go), it makes total sense to me to have as much
control as possible, as long as it©ˆs properly costed and charged for. This
is where a lot of people come unstuck....not charging for all the
additional work required to get from digital capture to press-ready file.

--
STUDIO LTD
Atelier für Fotografie
Altenbergstrasse 8
CH-3013 Bern

T: ++41(0)31 332 88 33
F: ++41 (0)31 331 62 42
M: ++41(0)79 227 02 19
U: www.studio-ltd.com
___________________________________________________________________________

Re: Rendering the print
Posted by: Ric Cohn
Fri Jul 13, 2007 3:23 am (PST)

If we're talking commercial photography I think the person who is most qualified should work on the Raw. IMO, this *should* be the photographer, but I also think that there is a real problem that some photographer's haven't taken the time to learn how to work with Raw to output files that can be handed off to a retoucher or printer without problems. In the days of film the most "professional" medium was transparencies, not negatives, and good photographers were experts in delivering transparencies that were properly exposed, with proper color and contrast for reproduction or further work by a retouching firm. If a transparency was bad it was usually immediately obvious and the photographer wasn't re-hired. Although it's also possible to have a transparency that looks great, but needs a lot of work to translate to print, and Printers frequently used this fact to blame the photographers, there was no Raw to go back to. Some clients have decided this isn't the case with digital photography and very unfortunately, once they get burned sometimes they decide they can get better results from the Raw than the photographers. Very very unfortunately, sometimes they are right.

Ric Cohn
___________________________________________________________________________
 
Delivering RAW files/was: Rendering the print
Posted by: "Andrew S. Webb"
Fri Jul 13, 2007 3:23 am (PST)

I didn't mean to suggest that photographers shouldn't do their own processing and post work. I was making a pitch recommendation for a potential "Digital Tech". I did say

"there are many photographers who have very little digital knowledge, and they're left adrift with a RAW file they don't know much about."

and lots of them have no interest or talent in digital processing and/or manipulation. They want to light and shoot and be done, because that is what they have always done. (i.e. Light, shoot, deliver chrome.)

If that ain't you, please don't take offense.

Cheers,

_andrew webb
___________________________________________________________________________

Delivering Raw Files
Posted by: "Chris Brown"
Thu Jul 12, 2007 4:48 pm (PST)

In my small world, here under this slab of concrete, I find the reason agency clients want the raw file is because they want to collect fees for image processing, retouching and CMYK file conversion from the advertiser. I've been fortunate enough to show my work & ability in that workflow (in part, thanks to Dan's class) and have been able to retain that income stream.

When I serve corporate clients, they've never wanted raw files. They simply want to drop the TIFF into their layout and be done. They appreciate the service of raw manipulation and CMYK separation, even with the added cost.

I recently did a job where the in-house A.D. was effusive in her thanks that I would deliver final, selected TIFF files ready for print. She was weary from getting 100+ JPEGs for each scene as final files from other photographers, which she then had to retouch and convert, a task she admitted she was not well-trained in.

Cheers ~
Chris Brown
___________________________________________________________________________

Re: Delivering Raw Files
Posted by: Dan Margulis
Fri Jul 13, 2007 7:01 am (PST)

Chris Brown writes (and Werner similarly),

In my small world, here under this slab of concrete, I find the
reason agency clients want the raw file is because they want to
collect fees for image processing, retouching and CMYK file
conversion from the advertiser. I've been fortunate enough to show my
work & ability in that workflow (in part, thanks to Dan's class) and
have been able to retain that income stream.

There's more than just the actual amount of money going into your pocket. You make yourself an important part of the operation in this way. Once you set up this workflow, you have made yourself difficult to replace, because photographers with your sklll level aren't particularly common.

Someone who supplies ONLY raw files is not contributing nearly as much skill, and IMHO is putting a big target on his own back. It would not be accurate to say that the client could hire a high school student to go out and shoot the scene and usually get the same quality results. But it *would* be accurate to say that the high school student could get equivalent results on *certain* jobs. And it would be accurate to say that clients often think to themselves about how to reduce costs, and that high school students do not cost as much as professional photographers.

I recently did a job where the in-house A.D. was effusive in her
thanks that I would deliver final, selected TIFF files ready for
print. She was weary from getting 100+ JPEGs for each scene as final
files from other photographers, which she then had to retouch and
convert, a task she admitted she was not well-trained in.

Right, which is another way of saying that she thinks the final result is likely to be better if you do it than if she does. And, of course, she has clients, too, who can notice these things. And if they are not satisfied with the work that she does, I doubt that she will admit to them that it is due to the fact that she is a poor retoucher. Rather, if she is like most art directors, she will blame it on poor photography.

Dan Margulis
___________________________________________________________________________

Re: Delivering Raw Files
Posted by: "Rick Gordon"
Fri Jul 13, 2007 4:07 pm (PST)

After having followed this thread, and a similar one on the ColorSync list, I'm still left with a sense that there are a lot of photographers out there whose digital skills are too limited to rely that a file that they have edited is not going to be damaged in some way.

There are also photographers who produce good content but don't have the basic photographic skill to provide optimal exposures, so their JPEGs may have clipping that might be retrieved in RAW.

I'd much rather have their DNGs (which this group is probably not generating) or raw files (including their XMP files, which I can reference and either use, modify, or ignore) to create print-ready work.

Other Raw-savvy photographers are purposefully increasing exposures to produce data that might have less shadow noise while retaining retrievable highlights, but will likely create a lousy JPEG. They are probably also skilled in manipulating RAW, so I probably won't need their Raws.

But to summarize, as someone who works in a context where I'm receiving content which is produced by photographers who have good cameras and good eyes but marginal digital technical skills (know any?), give me their Raws any day!

Rick Gordon

--

___________________________________________________

RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________________

WWW: http://www.shelterpub.com
___________________________________________________________________________
 
How JPEG Works (was: Rendering the print)
Posted by: Dan Margulis
Sat Jul 14, 2007 8:27 am (PST)

Henry Davis writes,

The mechanics and the reasoning behind how CR processing works would
be interesting but, I forgot to mention another part of the paper
that would also draw some attention: the mechanics and the reasoning
behind how in-camera JPG processing works. More hoopla and
discussion about in-camera JPG processing would be helpful to
photographers and Photoshop users with regard to technique and
decision making. Even people using cameras that can't make raw files
would be better served to understand how different camera settings
are designed to work.

In principle they are designed to make the process foolproof. The word "foolproof" is used advisedly. They are designed to make fools look like good photographers. And so, even though every camera model (not just every camera manufacturer) cooks up its JPEGs in different ways, they all have many features in common, because it is well known how to make images more palatable to fools.

Understanding the meaning of what the camera does before saving the JPEG depends on the following question. Suppose you have two possible methods of proceeding. Method One produces superior-looking images 80% of the time, in comparison to its opponent, Method Two, but in those cases the Method Two versions are correctable. 10% of the time Method Two looks better but Method One is correctable. 5% of the time Method Two looks better and Method One can only be corrected very tediously. And 5% of the time Method Two looks better and Method One is a total loss.

Given this description, which is the better method? Most professionals, I think, would like a choice of using Method One when it works better and Method Two when it doesn't. But we don't have a choice. The camera vendors choose one method or the other.

If we are forced to choose, then we professionals have to pick Method Two. We are willing to go through extra work if need be. We are not willing to have even a small fraction of images ruined. Unfortunately, the camera vendors feel that Method One makes more sense. If I were in their position, I'd agree. Their typical client is perhaps not a fool, but certainly not a very good photographer, who is apt to blame the ruined pictures on his own ineptitude rather than on an overly aggressive move by the camera's programmers. Their typical client will feel that if their camera produces better-looking images 80% of the time than the camera of some other vendor, then it is a better camera, when in fact it's all programming.

The secrets that all these algorithms employ don't vary much, only how aggressively they are implemented. The more aggressive, the better they do on typical images and the worse on atypical ones. The secrets are:

*Range is extended, usually to the max, even if the image is shot flat.

*Colors are brightened, particularly greens.

*If the lightest and darkest areas of the image seem to the camera to be likely to be white or black, the camera will force them to those neutral values.

*More contrast is assigned to the midtones and less to the highlights and shadows.

*The lightest and darkest parts of the capture are ignored, discarded, as it is assumed they are unimportant. The overall range is set using somewhat less extreme values as endpoints.

Of these items, the first is correct almost all the time.

The second is usually right, and if it isn't it is easy to fix *unless* the colors are already very pure, in which case there may be difficulty recovering detail.

The third is likely to be correct if applied conservatively, but the more aggressive the camera is, the more likely it is to produce an image with correct highlights and shadows and incorrect midtones. Such images are quite difficult to correct.

The fourth is more problematic. It's usually correct, but when not, detail in highlights and shadows gets hurt, and we may or may not be able to recover.

The fifth is the most likely to be correct, but the most disastrous when wrong. The idea that the lightest half percent or whatever of pixels is unimportant is correct at least 90 percent of the time. Even in images with critical highlight detail, that small percentage stands a good chance of not being missed. But every once in a while disaster will strike. The camera will have discarded information that is absolutely needed for quality reproduction and now we are in a fix.

In looking this group over, it becomes evident that the problems of camera- baked JPEGs are usually found where highlight detail is critical. Not just important--critical. Otherwise the JPEG will be fixable. The critical areas need not be limited to whites. They can also includes brilliant colors, where detail in the lightest channel is needed. Occasionally there is also an issue in shadows, but highlight problems are much more common.

The deficiencies of JPEGs are often wildly overstated by vendors who wish us to make more use of the raw format. We need to remember that the moves described above are in there because they all work well most of the time. When they do, there's no problem with using the JPEG unless you're doing something so stressful that it brings out artifacts of compression.

There should be no difficulty identifying the types of images in which the above automated approaches won't work well. Those are the kinds of images for which we *do * need to be able to access the raw file.

Dan Margulis
___________________________________________________________________________

Re: Delivering Raw Files
Posted by: "Chris Brown"
Sat Jul 14, 2007 8:27 am (PST)

there are a lot of photographers out there whose digital skills are
too limited to rely that a file that they have edited is not going
to be damaged in some way.

Rick,

Please define "damaged in some way".

Do you mean clipped? Skewed CMYK values? What do you mean?

Cheers ~
Chris Brown
___________________________________________________________________________

Re: Delivering Raw Files
Posted by: "Rick Gordon"
Sat Jul 14, 2007 11:13 am (PST)

I mean clipped -- either to black or white, or clipped in individual channels due to overzealous saturation adjustments, or badly sharpened.

I'm less concerned with skewed values. I frequently get images that have overall casts that are not defendable as an atmospheric effect, but that's fixable.

The ability to use a photographer-corrected file (and of course, his/her monitor is reasonably calibrated, right?) for reference AND a raw is ideal.

Again, I'm not referring to material received from skilled correction artists, but rather from the many amateur or semi-pro photographers, or pros without well-developed digital skills, who supply images for publication.

Rick Gordon
___________________________________________________

RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________________

WWW: http://www.shelterpub.com

------------------------------------------------------------------------
___________________________________________________________________________

Do Your Curves Throw You a Curve?
Posted by: Stephen Best
Tue Jul 24, 2007 9:22 pm (PST)

Available for download here:

http://luminous-landscape.com/pdf/Curves.pdf

I just read this and it all seems to reinforce Dan's point that both the composite RGB curve in Photoshop and Camera Raw affect global saturation. Pulling back saturation after a composite move (Figure 23) just seems crazy to me.

Stephen Best
Macquarie Editions
___________________________________________________________________________

Re: Do Your CurvesThrow You a Curve?
Posted by: Stephen Marsh
Wed Jul 25, 2007 9:18 am (PST)

The source link for Mark Segal's PDF can be found here:

http://www.luminous-landscape.com/essays/Curves.shtml

Stephen Best wrote:

I just read this and it all seems to reinforce Dan's point that both
the composite RGB curve in Photoshop and Camera Raw affect
global saturation.

It may do that, but it appears to be written with a rather different intent and counters many of Dan's criticisms. I mean that in a positive way, as it is good to know the issues and how to work with/around them.

The article does exclude detail in mildly/higher saturated red images. So perhaps one conclusion is that with the various "counter-moves" in the RAW converter/renderer to combat the saturation increase in certain images that kills critical detail - that it may be possible to achieve the desired results directly in ACR/ALR. For some other images, notably detail in critical red hues - it is advised that the image be corrected in Photoshop or another CR package that does offer luminosity or separate R, G, B curve edits as opposed to the saturation increase inherent in ACR/ALR tonal moves that may not be successfully countered as with other hues.

For me, two important things are gained from this on-going discussion.

i) We know to be on guard for loss of detail in red hues, above and beyond others. Are green and blue also problematic too, or is it only red? From past topics on saturation and yellow hues, we know that there are good (smooth, natural) and bad (posterized, artifacts) ways to increase yellow saturation. It is important to the image editing communities knowledge pool to know when to pay extra attention for certain classes of images and how to handle them in certain situations.

ii) Future versions of ACR/ALR may offer the "simpler" luminosity curve option as a checkbox or other control and or gain separate channel or LAB capabilites (as offered in competing products). This is good for end users.

As a result of such discussion and feedback/requests to Adobe, and if Adobe share the end users vision - we may see such features in future releases of ACR/ALR.

It is also subjective, Dan and Mark are evaluating different images (Mark was not looking at detailed red hues, he made that point clear and agreed with Dan's position in this one case) and Dan was writing about an earlier version of ACR than used in the article (we have been told many times by up to date ACR users how outdated older versions of ACR are and that they should not be compared to each other).

Pulling back saturation after a
composite move (Figure 23) just seems crazy to me.

Perhaps, but an easier/quicker hoop to jump through for many than separate curves to each image channel (a flaming hoop perhaps?)--and one that is natural for an ACR/ALR user that does not have access to individual curves anyway.

Regards,

Stephen Marsh.
___________________________________________________________________________

Re: Do Your Curves Throw You a Curve
Posted by: Stephen Best
Wed Jul 25, 2007 6:55 pm (PST)

Pushing saturation to the gamut boundaries will kill information (colour variation) that can't be reclaimed by a subsequent reduction. It's something you may get away with (depending
on the image colours and space) but it's hardly an ideal practice.

I fought this whole "correct with composite curves then correct for the unwanted hue/ saturation shifts" rigmarole for years before I stumbled onto separate luminosity and colour layers (all in RGB and as outlined in an earlier post of mine here). I surprised at the lack of interest in such an approach or are old practices so entrenched?

Stephen Best
Macquarie Editions
___________________________________________________________________________

Re: Do Your Curves Throw You a Curve?
Posted by: Dan Margulis
Thu Jul 26, 2007 10:19 am (PST)

Stephen Marsh writes,

For me, two important things are gained from this on-going discussion.
i) We know to be on guard for loss of detail in red hues, above and
beyond others. Are green and blue also problematic too, or is it only
red?

In principle they are also problematic (greens even more so than reds) but in practice reds are the more difficult case. The most serious problem is provoked by an area with pronounced color where the lightest channel has important detail that extends into the highlight. Many red objects answer this description. In that case, the ACR range-setting commands suppress detail in the weak channel. As that weak channel is critical to detailing in deep colors, the ACR routine can damage the image beyond repair, as we've seen.

Green objects seldom have this attribute. We see plenty of green things, but not vividly green in the sense that the poppies shown in PP5E or the cherries that we worked with on this list are vividly red. Instead, the weak (green) channel usually lives in the quarter- to mid-tone. Now the problem is not so much a wipeout of detail as it is that contrast is misallocated and will have to be recorrected later in Photoshop.

Blue objects that answer this description are apt to be clear, deep skies. Detail in these is usually not important so it does not matter whether it gets lost.

It is also subjective, Dan and Mark are evaluating different images
(Mark was not looking at detailed red hues, he made that point clear and
agreed with Dan's position in this one case) and Dan was writing
about an earlier version of ACR than used in the article (we have been
told many times by up to date ACR users how outdated older versions of
ACR are and that they should not be compared to each other).

This is all correct. In an earlier post to Mark, I cited half a dozen specific images that create problems for the ACR implementation. None have been challenged by Mark. Jeff Schewe took a shot at one of them. He could not come even close to an acceptable result using Photoshop CS2. With the additional capabilities found in CS3, he was able to get closer but still not compensate for the detail loss. As I understand it, Mark suggests in his paper that such red objects are better handled elsewhere.

Mark, for his part, shows half a dozen images that do *not* create problems for ACR. These aren't challenged, either. He's right--ACR would handle them adequately.

The deficiencies of the ACR approach are most apparent in images with one of the following three characteristics:

1) They contain large, important areas of pronounced color that also carry important detail.

2) They have conflicting casts due to unusual lighting or reflections.

3) They have a single uniform cast, but also important d etail that resides in several ranges.

All of the images I cited have one or more of these characteristics. *None* of the images Mark shows have them. They feature dull colors. Where colors are dull, the channels behave more or less alike and a master curve is almost as good as channel-by-channel. Objects that are more strongly colored fall in significantly different areas of each channel. Any operation that treats all three channels alike necessarily damages such images.

The only brightly-colored large objects I saw in Mark's piece were Color Checker swatches. As swatches contain no detail, there is no visible damage.

Mark shows AFAIK no example of an image with conflicting casts. His example of an image that has a single cast does not have detail in all ranges; everything of importance is in light areas. Add a person's face to the image and it the flaws of the ACR approach become clearer, as the similar example shown in PP5E demonstrates.

It seems to me that the question is not how many images are *not* damaged by extending ranges in ACR as to whether there are any that *are*. To say that an approach that boosts saturation without user control should be mandatory on the grounds that it is acceptable for the majority of images (which I don't understand Mark to be saying, but others have) is like saying that the auto-focus and auto-exposure settings on a camera should be made mandatory. They're harmless most of the time, too.

The inadequacies of the master-curve approach being used in ACR's range-setting routines have been known for twenty years or more. Stephen is correct in pointing out that ACR lags the competition here even though it surpasses it in several other areas. Unfortunately, range-setting is more critical to image quality than the more glamorous bells and whistles are. Stephen may be right that on the whole nonexperts may now be able to get better results from the relatively intuitive controls in ACR, in spite of the deficiencies in range-setting, than by trying to attack images in Photoshop.  

This is not, however, the audience that concerns me. For that other audience, David Marley, who is an expert retoucher, is correct. "If you, as a professional photographer, prefer to sacrifice detail for the convenience of ACR's controls, that is your choice. Of course, this assumes you are aware of the full impact of these settings. "

Dan Margulis
___________________________________________________________________________

Re: Do Your Curves Throw You a Curve?
Posted by: Stephen Marsh
Sun Jul 29, 2007 6:42 am (PST)

Dan Margulis wrote:

In principle they are also problematic (greens even more so than
reds) but in practice reds are the more difficult case. The most serious problem is
provoked by an area with pronounced color where the lightest channel
has important detail that extends into the highlight. Many red objects
answer this description. In that case, the ACR range-setting commands suppress
detail in the weak channel. As that weak channel is critical to
detailing in deep colors, the ACR routine can damage the image beyond
repair, as we've seen.

OK, thanks Dan, I thought that this may have been the case from the past threads on yellow hues and the associated problems of increasing saturation using some methods over others. RGBCMY hue images with detail may have general and unique issues to take into account when editing, some more so than others. Preserving visible luminosity detail and variation can be problematic in both RGB and CMYK, not to mention preserving minor hue/saturation variation/detail.

I look forward to CR file examples from list members that can help us to examine and combat this problem. Even though it should not matter whether the data being processed is CR or not, it would be good to have CR file examples. I have a feeling that the images most affected by this issue will not be common or typical images for most folk, but could be a common issue for certain subsets of folk editing images(product shots, still life, landscape etc). And no – for those that will no doubt ask, I am not going to define what a typical or common image is.

Green objects seldom have this attribute. We see plenty of green things, but
not vividly green in the sense that the poppies shown in PP5E or the cherries
that we worked with on this list are vividly red. Instead, the weak (green)
channel usually lives in the quarter- to mid-tone. Now the problem is not so
much a wipeout of detail as it is that contrast is misallocated and will have to
be recorrected later in Photoshop.

I think that you are referring to natural greens here, and not those found in product shots for example, where one may encounter more unique editing issues.

Blue objects that answer this description are apt to be clear, deep skies.
Detail in these is usually not important so it does not matter whether it gets
lost.

Fair enough, so it is good to know that RGBCMY hues in certain images can present issues with luminosity variation in RGB editing, let alone conversion to CMYK.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: Do Your Curves Throw You a Curve?
Posted by: Dan Margulis
Sun Jul 29, 2007 5:00 pm (PST)

Jeff Schewe writes,

To be accurate, Dan, I never tried opening your red sample image in Camera
Raw 3 in Photoshop CS2...couldn't since the ability to even OPEN JPEG images
wasn't added until Camera Raw 4 in CS3. I never opened your images in
CS2-period.

I apologize for the sloppy language. You did indeed work in CS3, however in your first effort you used only commands that are also found in CS2 and could thus have been duplicated there on a suitable raw file. In your second effort, you employed Split Tone, a command not found in CS2. Therefore, this second, improved effort would not have been possible in CS2.

And I did indeed "compensate for the detail loss" as noted by several other
members of the list. No, it was not numerically a match but to my eyes, it
WAS pretty close, particularly after you pointed out I lost saturation in
the shadows and used split tone to bring them back.

It was a whole lot closer, I'll give you that, but the detail loss was still present. Certainly it was close enough that possibly additional tweaking in ACR might make it even better, or if not it could be fixed up easily enough in Photoshop, thus demonstrating that with five steps in ACR plus an additional move in Photoshop we can probably equal the quality of what most camera vendors' acquisition packages would give us in a single set of channel curves.

Be that as it may, for the purpose of this thread the significance is that (and I hope we now agree on this point) the desired correction could not have been accomplished in CS2. My book makes clear that I have studied CS2, the then-current version, only. I have reminded the list several times of this, and also that I predicted that future versions of the module would improve rapidly. I have looked at CS3 but not tested enough to make generalizations about what it can or can't do.

This thread began when I stated that, contrary to the conventional wisdom, ACR's range-opening routines employ, in effect, a master curve, resulting in increased saturation as the image lightens, in a way that the operator cannot control. I stated that this damaged certain images in ways that I had demonstrated in PP5E with specific images, cautioning that that book was written for CS2 and not CS3.

These statements caused such a flood of vituperative, personally abusive statements accusing me of not knowing what I was talking about that the acting list moderator took drastic action against uncivil commentary.

Since then, you have conceded that the ACR routine *does* increase saturation. It goes without saying, I think, that this can cost detail in already saturated images such as the cherries and other things of similar color. The master-curve behavior of operating on RGB channels prior to making other adjustments has been clearly shown by reference to the individual channels.

The only thing, therefore, that is not been confirmed is that actual images exist that are damaged if ranges are opened in ACR. I have offered two such images in my book, made the raw files available to readers, showed the damage in comparison to if the range had been opened in a normal way in Photoshop, asserted that it could not be repaired either within or outside of ACR in CS2, and suggested that this a strong argument in favor of opening ACR files conservatively.

One of these images is a picture of red poppies--a raw file similar in conception to the cherries image, with similar problems but even more pronounced detail loss. The damage is apparent when comparing the version with range opened in ACR as opposed to one acquired simply in ACR with range being extended with Photoshop curves.

If that damage can be compensated for within CS3, it's a useful piece of information that many would be interested in knowing. If, however, the objective is to show that I was wrong in what I said, one can't reasonably use tools that weren't available at the time I said it.

This poppies image has been tried by several beta readers, who confirmed my results. Based on your experience with the cherries image, I believe you would not be able to correct the damage either, if limited to CS2, for the same reasons: you would be stuck with either damaging the detail in the lighter parts, or losing saturation in the darker areas.

And all this was done on a JPEG in sRGB mind you. . .not a raw image for
which Camera Raw has been designed...

That's right. You have brought up this point before, so let's revisit why it was done. You entered the thread, and indicated that the commands were luminosity-based and did not result in an increase in saturation, citing Bruce Fraser and Real World Camera Raw as a source of the information. In response, I suggested that you or other list members could test the matter for themselves with any image containing strongly detailed, vivid red objects. I explained how, in such an image, one could determine whether the curve being applied was luminosity-based, or was based on a master curve approach.

Although I recommended using one's own raw image, I provided a JPEG of the cherries image. The reason I could not provide a raw was that the site I was posting to does not support raw format. Although a JPEG, it is suitable for evaluating what the commands do. After considering this and consulting with your friends at Adobe, you acknowledged that there was, in fact, a saturation boost and stated that you would alter the text of the next edition of RWCR to reflect this.

The only problem with this JPEG is in evaluating your latest (CS3) correction. There, you were using other commands as damage control to reverse what the CS2-based commands did. In some circumstances, this can create disagreeable artifacting. We'll never know whether that happened here, because the inherent JPEG artifacting will hide it if it's there.

I still don't have access to any other raw examples where you claim CR does
an unprofessional job...

According to list records you were actively participating on the day when I posted the list of images and described specific problems, so you may have overlooked it. The listing is at
http: //tech.groups.yahoo.com/group/colortheory/message/17990

For a full explanation of what is desired from the raws that are on the PP5E CD, you would need a copy of the book, which Peachpit will doubtless send you if you need it. That will show the desired result and the methods you would be competing against. While we can exchange files ourselves and with others who already have a copy of the raw file, the file itself is not permissioned for public posting, unlike the cherries image. John Ruttenberg has given us permission to repost his ballet image, which is another raw file on the list. I'd predict, without guaranteeing anything, that you can't get a good result in CS3. With respect to the two images from PP5E, I am guaranteeing that you can't get a good result in CS2, feel free to try to prove me wrong, but not making any prediction about CS3.

If you'd prefer to work with your own raw images, in
http: //tech.groups.yahoo.com/group/colortheory/message/18270
I listed the categories of image where problems can be expected in ACR. The ones Mark showed in his paper don't fall into any of these categories, but you certainly must have a lot that do. If you'd like to make some available for testing, it might be wise for us to exchange them offline before working on them so that there wouldn't be any later dispute over whether they were appropriate images.

If I have ever said that CR is an unprofessional program, I should not have. My comments are aimed only at its methods of establishing range, which are primitive in relation to the rest of the product. Some may find them satisfactory; high-end retouchers would not, for a variety of reasons. The images I've cited have glaring deficiencies. In many other cases the problems look small in the original but can adversely affect final quality.

Many list members are aware that Jeff has agreed to assume authorship of the semiofficial guide Real World Camera Raw, replacing the late Bruce Fraser. These are big shoes to fill. I know the list wishes Jeff well as he faces deadlines in preparing the CS3 version of this title.

Dan Margulis
___________________________________________________________________________

Moderation Notice
Posted by: Dan Margulis
Sun Jul 29, 2007 4:59 pm (PST)

I will be vacationing beginning tomorrow. Stephen, Darren, and John will be monitoring posts.

We regard the "Do Your Curves Throw You a Curve" thread as being a continuation of the previous "Rendering the Print" thread. A warning to the list with respect to civility in this thread remains in effect, along with the "Color Management. Personal attacks, unprofessional language, or the like will result in rejection of the post at the discretion of the moderator on duty.

We continue this warning not because of any recent problem here but because this topic recently emerged on at least three other lists, two of which promptly degenerated into cesspools of namecalling.

I will return next Tuesday, after much time on the beach exercising my elbow.

Dan Margulis
___________________________________________________________________________
 
Re: Do Your Curves Throw You a Curve?
Posted by: Stephen Marsh
Mon Jul 30, 2007 5:00 pm (PST)

After reading the thread at Luminous Landscape that is "discussing" the PDF in question, I am glad to see that it is not just this list that has the ability to run off topic.

Something that is worth commenting on briefly, just in case list members here misunderstand the preference for lightness curves with no saturation boost associated with them - this is a step to the final image, not the final image edit itself. Yes, it is agreed that moderate isolated luminosity moves are often unflattering on many colour images - the colour component also needs a boost too. The Shadow/Highlight command defaults to a moderate boost in saturation (strangely named "colour correction"). I have no opinion if this is from being familiar with film response or if it is part of inherent human perception or whatever, I am sure there is a study somewhere that can be cited (I think Dan went into this with the Tiger in the grass/water image in his Professional Photoshop books).

It is a minor step to blend in the better detail from a luminance edit over a more saturated image, but it can be harder to recover lost detail that may be hosed in highly saturated areas. If one can counter the built in saturation increase with negative saturation/vibrance, clarity or whatever and retain both saturation and detail when required, then great! If one can do the two separate renders to colour and luminosity detail in ACR/LR and then merge them in Photoshop with no tonal edits in Photoshop, then no harm no foul. If one has to do part of this in Photoshop and part in ACR/LR, then that is OK too. Just as one may blend two exposures, one may blend an image that is good for detail into one that is good for colour (or vice versa), in either RGB or LAB depending on taste or ideology.

This point is of course independent of the ongoing debate about composite RGB curves, ACR/LR lightness/saturation increase in various tonal controls (basic adjust and curves tabs), individual RGB channel curves and L channel curves or simulated luminosity blend curves in RGB and how detail may or may not be affected, in gamma corrected rendered files or linear camera raw files.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: Range setting confusion
Posted by: Lee Clawson
Sun Aug 12, 2007 10:08 am (PST)

David,

If you are interested in this question, that is, the impact of Adobe Camera Raw on detail in saturated reds Mark Segal recently published his findings on Luminous Landscape. The article with examples of photographs having strong predominant colors is posted at:
http://www.luminous-landscape.com/pdf/Curves2.pdf.

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: Range setting confusion
Posted by: Dan Margulis
Mon Aug 13, 2007 7:46 am (PST)

Lee Clawson writes,

David,
If you are interested in this question, that is, the impact of Adobe Camera
Raw on detail in saturated reds Mark Segal recently published his findings
on Luminous Landscape.

There are many problems with the way this paper was presented. First, when I refer to specific images, they are made available. Two members of this list, John Ruttenberg and Ric Cohn, posted their raws, and two others are available to any owner of PP5E. Therefore, nobody needs to take my word for what images show--they can open the files and see for themselves. Here, paper, all we have is some postage-size images in a low-res PDF, and we are asked to take the author's word for their meaning.

Second, when I show an image, I show the alternatives as best I can. The first paragraph of Mark's original paper indicates that he will do so as well. It states: "The purpose of this essay is to explore whether one achieves superior image quality by [A] maximizing the use of White Balance and luminosity (contrast and brightness) adjustments in Adobe Photoshop Camera Raw (CR), followed by further adjusting luminosity of the rendered image in Photoshop (PS) using the RGB composite Curve (PS-RGBc), or instead [B] rendering the raw image with little adjustment in CR and using the PS individual channel curves to adjust White Balance and luminosity."

Nothing like this is forthcoming in either paper. The only time Mark explores Option B, he doesn't use channel curves--instead, he compares ACR to the master curve of Photoshop, with predictable results.

I did not challenge any of Mark's first paper, because Camera Raw is harmless enough in images of the type he was using as examples. Instead, I named three categories of image that *are* problematic.

As to the second category, images with conflicting casts, Mark concedes that "CR is not configured to resolve this unaided by Photoshop." Although other raw modules have no problem with such images, fair enough.

Mark says that the third category, images with a single uniform cast but important detail that resides in several ranges, "isn't worth evaluating further, because it describes routine conditions." Since he does not back this up with an actual image, there is nothing more to say.

Mark does address the first category, images containing important objects of pronounced, saturated color with important detail, with three examples. It is easiest to take them in reverse order.

The third image, a red flower, is a nonstarter because the method shown depends on use of the Vibrance command, which is found only in CS3. As indicated several times both in my text and in this thread, my analysis was of CS2 only. If one wants to use this example in a forward-looking sense, saying that CS3 adds possibilities that were not present in CS2, fine. But if the purpose of the paper is to criticize what I wrote, the use of this command invalidates the criticism.

This has already come up once on this list, when Jeff Schewe was working on the cherries image that I had posted. Using only tools that were found in CS2, he was unable to get a satisfactory result. He then improved upon it by using Split Tone, a command found only in CS3, and I immediately objected in
http: //tech.groups.yahoo.com/group/colortheory/message/18297

I said, "The damage is apparent when comparing the version with range opened in ACR as opposed to one acquired simply in ACR with range being extended with Photoshop curves.

"If that damage can be compensated for within CS3, it's a useful piece of information that many would be interested in knowing. If, however, the objective is to show that I was wrong in what I said, one can't reasonably use tools that weren't available at the time I said it."

For these reasons, Mark's third example is irrelevant On to the second image, which *is* a fair example. It offers a red flower of the type that I showed in PP5E, and it limits the correction to methods available in CS2. Even though we have only the thumbnail PDF images to look at, it is clear from comparing them that Mark's final result is unacceptable, for the same reason that Jeff's first attempt to correct the cherries image was unacceptable--it just trades one problem for another.

Mark shows the original, uncorrected image in his Figure 21. Plainly, it is too flat, since nobody has opened range yet. But doing so in ACR, as we all should know by now, blows out critical detail in the lightest channel, in this case the red. Mark shows and comments on this loss of detail in Figure 22. The color is brighter and more pleasing, but professionals would never accept it because of the detail loss.

In CS2, the only available countermeasure is a blunderbuss desaturation of all reds. That's what Mark does in Figure 23, his final version, as Jeff did in his first try to correct the cherries. One simply cannot do that to a picture that is supposed to portray vivid colors, as this flower should. It puts the detail back into the lighter areas at the cost of muting the overall color. The final result is too gray to be acceptable. What the client would want is something as colorful as Figure 22 while retaining the detail of Figure 21. That result is available in any program that contains channel curves, but not in ACR as of CS2.

On to the first image, which is a still life of oranges and pink grapefruits in a wicker basket. In addition to the question of loss of detail in the colored areas, Mark notes that it arrives with a cast and suggests that it answers the description of category #3 above. It does not. Oranges and pink grapefruits are both light objects, as is the basket. Nothing of importance occurs in the midrange. If a lime, a pear, an apple, and red grapes were added, then the image would qualify.

Unlike the second image, where there was enough to compare between the thumbnail images to know that the final result was unacceptable, here it would be difficult to say how bad the image is. It's pretty hard to see detail in yellows. I would be very dubious that this is an acceptable result but without seeing the actual file I could not say for sure.

It is helpful, though, that Mark showed the progress of two channels as he applied steeper and steeper S-shaped curves to the original. In the cherries image of a few weeks ago, I explained how ACR uses a master-curve approach to setting range. That is, it applies a single, uniform correction to each channel, then averages them, performs other calculations, and regenerates the final RGB document. This is a naive approach. Obviously, ACR should contain channel curves as every professional-level program does. But if there is only a single curve, then the correct approach is to average the channels together before applying it, instead of applying the same correction to the channels separately. Then, saturation could be added in a variety of ways--including, if desired, as an automatic part of the curving process.

If there was any doubt that this was the case from the cherries image, Mark's fruits dispel it. In the red channel, which is the lightest, if an S curve is applied as a master curve, contrast in the fruits will be reduced, but contrast in the wicker, which would fall in the steep part of the curve, would be increased. Similarly, the blue. Shadow contrast would decrease, but the wicker contrast would increase.

Mark's examples show that this is exactly what is happening. If the channel data had been averaged before applying the curve, this would not occur: every channel would gain contrast in both the fruits and the wicker.

The steeper the center of the S curve, the worse it is for the red channel. Mark correctly notes that nevertheless overall contrast does not appear to decrease as the curve gets more pronounced, but his analysis of why is not correct.

Overall contrast loss is avoided because the green channel, which is not pictured, falls in the steep part of the master curve. As the green channel is around twice as important as the red, and six times as important as the blue, any improvement in it will cancel out or perhaps even outweigh any deficiencies in the other two channels. Nevertheless, there is no reason to settle for a bad red channel (the blue is bad too, but unimportant). Therefore, stressing again that I have not seen the actual image but only the thumbnails, it seems unlikely that this is a satisfactory result. (The ACT course offers a similar exercise with an image of a lime, so I'm very familiar with what type of channel structure results in a good image--I've seen hundreds of people work on it.)

The inadequacy of this result could be demonstrated by comparing it to an image properly corrected in Photoshop, which Mark purports to do--except that he uses a master curve! So, of course, it damages the image in the same way that ACR does. In something as orange as these objects, the red channel is light, the blue is dark, the green is in the middle. These are short ranges for the purpose of channel curves--easy to emphasize detail. But for the purpose of a master curve, every tonal value from lightest to darkest is important in * some* channel. By definition, a master curve of any shape *must* damage at least one channel in an image of this type. Nobody with significant curving experience would work in Photoshop this way.

Certainly Mark is to be commended for having put these papers together and for suggesting ways in which people can work in ACR and ciscussing some of the alternatives within the modules are. However, Because there is no valid comparison with other methods, and no way for readers to make the comparison themselves, the sum of what Mark's two papers demonstrate is that it is possible to apply certain corrections in Camera Raw that make an image look better than if they had not been applied. It is unclear that 40+ pages are required to make this point successfully.

Now that the command has been improved in CS3, one could write a similar paper arguing that one should neither use ACR nor channel curves, but rather Brightness/Contrast. Images might be shown of elephants and the like, and it would be easy to see the improvement that Brightness/Contrast made, just as one can see the improvements in Mark's ACR images. In fact, using a combination of Brightness/Contrast on certain channels and Levels on others would likely get better results than Mark did on the first two images, because we would have the option of leaving some channel(s) alone if the commands would damage them.

Dan Margulis
___________________________________________________________________________

ACR Image Processing Pipeline Info Requested
Posted by: Stephen Marsh
Thu Aug 16, 2007 6:55 am (PST)

With all the talk of camera raw editing, I would appreciate it if the basic ACR/ALR image processing path was explained in point and expanded form. I have not seen a good simple explanation yet - and I do try to stay informed (but being in prepress this is more of a hobby than in relation to my day to day duties). This is always skipped in any ACR discussion, perhaps because it is presumed basic and known to all? I am the first to admit that the mechanics of CR are not my main area (many photographers love the 'geek aspect' and thus take to CR like they did to film, lenses, lighting etc).

Most of the talk centres around cameras with a Bayer or similar colour filter pattern. Does ACR support the Foveon? How does this change raw image processing? What about scan backs? Can one access raw sensor data and if so how does this change raw image processing? Are the white balance and exposure processing stages the same as raw gray?

Scanners are not the same as digital cameras, but they are not far removed either. The flatbed scanner at my day job has a very linear tone response. Why was there no push for linear raw scanner sensor data editing? There was a market if there was a benefit (and perhaps even if not with good marketing), before digicams scanners and scan software was a big issue, that seems forgotten today with digicams being so common.

A general example of what I am looking for can be found here, although I seek to know the Adobe specific steps in ACR/ALR:

http: //www.xs4all.nl/~tindeman/raw/color_reproduction.html#_Toc116325711
http://www.xs4all.nl/~tindeman/raw/img/pipeline.png

http://web.cecs.pdx.edu/~cklin/demosaic/

By requesting a simple, clear explanation of the image processing steps taking place I hope to better understand the edits taking place. There are a lot of gray areas in this topic (pun intended).

We start with camera raw sensor data, grayscale pixel information in a linear state representing light levels captured. The ACR editing process is metadata driven, but one still has pixels. The pixels start off gray, then colour is created from the gray. This has to happen early in the process. I presume white balance and exposure come next. I am not talking of what order the user may move the sliders, but the order that is executed by the rendering engine.

We have the demosaic process where coloured pixels are interpolated from the gray pixels. In what colour mode, XYZ or RGB?

We also have the internal camera raw RGB editing space of ACR/ALR (custom ProPhoto RGB with altered 1.00 gamma).
 
Finally there is output of the rendering to a gamma corrected RGB editing space and bit depth selection.

In between there are adjustment steps such as white balance, exposure compensation, brightness, contrast, blacks, tone curves, etc. What processing is done in XYZ (if any) and what in RGB? I suspect more processing is done in RGB. In what mode is sharpening done? I suspect RGB, but we are also told that only gray levels are sharpened. So I presume math similar to the luminosity blend mode in Photoshop is being used to sharpen all three RGB channels the same, or is another calculation such as RGB>HSI performed and the calculated I channel component sharpened before being returned to internal editing RGB?

For years we have been warned about converting to or editing gamma corrected RGB files in a linear gamma RGB editing space (with one [in]famous exception endorsing this workflow). Then along comes camera raw data, which is linear by nature and thus linear editing is appropriate in this setting. As ACR can now open non camera raw files and also save them as DNG files one should keep in mind that there would appear to be a conversion to the gamma 1 internal editing space for editing/storage, if I understand things correctly. Again knowing the nuts and bolts would help in making an educated and informed decision about the use of the various features offered in ACR.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Mark Segal
Thu Aug 16, 2007 11:07 am (PST)

Stephen,

Several books will address most of what you have asked for. Start with Bruce Fraser's book on Camera Raw for CS2 (this is the second edition of his Real World Camera Raw). This will not carry you into the major changes which appear in Camera Raw 4.1. But the Lightroom develop engine is essentially the same as CR 4.1 (except it has niftier user interface features). Martin Evening and Scott Kelby have both written very good books explaining the Lightroom workflow. Michael Reichmann and Jeff Schewe produced an excellent video download on Lightroom, which you can purchase through the Luminous-Landscape Store (www.luminous-landscape.com). If you can wait a while, you can pre-order from Amazon.com Jeff Schewe's forthcoming book "Real World Camera Raw" which appears to be scheduled for some time in October. This will cover all the new CR features in 4.1.

Mark Segal
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: André Dumas
Thu Aug 16, 2007 3:47 pm (PST)

Hello Stephen,

Your question: "Why was there no push for linear raw scanner sensor data editing? " is an interesting one. I've been asking myself that very same question. Could it be because scanning conditions are so much more stable than photographing conditions ? Or that we can always rescan until we get the perfect scan ? Anyone knows ?

It seems to me that having a "raw scan", which would make re-purposing possible, would be a good thing.

André Dumas
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Chris Murphy
Fri Aug 17, 2007 3:44 am (PST)

On Aug 16, 2007, at 2:53 PM, colorman042000 wrote:

It seems to me that having a "raw scan", which would make re-
purposing possible, would be a good thing.

a.) It implies scanning film. Film is already rendered. It's inherently not the same thing as a raw capture. A scan is a 2nd capture of a scene. It's not the original capture. So scanners and cameras are actually really different. They have a sensor technology in common, but not much else.

b.) There is a single light source with a scanner and the primaries used are based on film and light source. Scanners are ripe with metamerism issues (when you start to scan things that aren't made from film).

c.) Handing over the scanner data stream to the end user isn't something I've heard users ask for, let alone scanner vendors at the time thinking their customers would have ever wanted. They prided themselves on their proprietary software, which was rather inextricably linked with the hardware, to process that stream and allow it to be manipulated.

As for Foveon and camera backs, those shots are considered rendered. I'm not sure if the software for those cameras has direct access to linear encoded, non-rendered data. It would then be considered "raw" the one thing unneeded with them is demosaicing.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________
 
Re: Range setting confusion
Posted by: Mark Segal
Thu Aug 16, 2007 6:56 am (PST)

I should apologize to readers for the length of the discussion below, but Dan has raised so many points that it just took a while to address them. By way of preface, my only reason for writing the two papers is that having done so much work in CR, I found myself fundamentally disagreeing with Dan's basic views about using CR for adjusting image contrast, so I implemented a number of tests to determine whether or not I was fooling myself about the quality of this program. I determined that I was not, and I decided it was research worth sharing. This is the second time Dan questions the usefulness of my results on the grounds that the images and procedures don't reveal the program's defects. After carefully reading his commentary, I have come to the conclusion that major parts of Dan's reply don't address the issues my paper addresses, so we're kind of talking past each other. The only way to cut through this is item-by-item, so here we go.

There are many problems with the way this paper was presented. First, when I
refer to specific images, they are made available. ....Therefore, nobody needs to take my word for what images show--they can open the files and see for themselves.

I never said my raw images are not available. Interestingly, Dan did not ask for them before he issued his critique of my paper or his comment that my files are not available; this has consequences explained below. It is correct that some List members have granted access to some images. To use the specific files Dan worked-up in PP5E does require buying the book. Anyone who wishes to probe my results may ask for my raw files and the raw file from Andre Dumas which I used, provided their email accounts or FTP servers can handle 10MB each of data. Alternatively, I could check whether I can up-load these files to the "Files" section of ACTL.

Here, .. all we have is some postage-size images in a low-res PDF, and we are
asked to take the author's word for their meaning.

Michael Reichmann and I agreed on the PDF format of presentation for practical reasons given the length of the articles and the time involved setting it all up in html. Looked at on a properly calibrated and profiled monitor, the PDF images show the points being made very adequately, and most are not postage-stamp size images. Many of them consume half a page so people could properly see the effects I describe. Viewing such images on a monitor reveals at least as much as they are intended to demonstrate compared with viewing images in a well-printed book. Therefore I think this is a non-issue.

Second, when I show an image, I show the alternatives as best I can. The first
paragraph of Mark's original paper indicates that he will do so as well. ...
Nothing like this is forthcoming in either paper. The only time Mark explores
Option B, he doesn't use channel curves--instead, he compares ACR to the master curve of Photoshop, with predictable results.

Where relevant I compared techniques using CR and Photoshop. When I use PS I use the composite RGB Curve for adjusting Luminosity, because I am trying to compare in a consistent manner and there are no separate channel curves in CR to be compared with. In PS, I would use separate channels for WB unless I do this in CR, which is technically preferable; with WB in CR, colour correcting a single cast with individual channels in PS is not necessary..

named three categories of image that *are* problematic....
As to the second category, images with conflicting casts, Mark concedes that "CR is not configured to resolve this unaided by Photoshop." Although other raw modules have no problem with such images, fair enough.

Dan was asked on ACTL to name these other raw converters; as well, if he says they are superior, he should demonstrate it. I have not sighted any such information. If there are any readers who have tested the use of curves functions in a number of raw converters with and without individual RGB curves for the specific task of remedying conflicting colour casts, it could be interesting if they would share their technical experience and results here.

The third image, a red flower, is a nonstarter because the method shown depends on use of the Vibrance command, which is found only in CS3. As indicated several times both in my text and in this thread, my analysis was of CS2 only. If one wants to use this example in a forward-looking sense, saying that CS3 adds possibilities that were not present in CS2, fine. But if the purpose of the paper is to criticize what I wrote, the use of this command invalidates the criticism.

The CR4 toolset is a package whose various elements can and should be used in combination to achieve the intended effects on contrast, brightness and saturation. Therefore it is perfectly reasonable to do so, and images which benefit from this flexibility are not "non-starters". Specifically, if Vibrance used in combination with a tone curve produces a more desirable combination of contrast and saturation, both are deployed for optimizing the two variables together, because the nub of the hypothesized damage problem is the hypothesized chain from a composite Curve contrast boost to excess saturation to loss of detail. As I was evaluating the use of CR4 for adjusting these images, it is appropriate that I use this version of the application. It has been publicly available first as a beta then a commercial release for about eight months now, enough time for anyone to have evaluated its merits quite thoroughly. I've processed over a thousand images with it, encompassing a wide variety of conditions. I think I know the strengths and limitations of this application in some detail.

As I've told Dan well before he posted his response to Annex 3, my papers are forward-looking, hence he knew that before he questioned my context. So why is he doing so now? I also told Dan at the same time that I'm not criticizing his book, he won't find any criticism of his book in my papers, and I would not be silly enough to criticize not using something that didn't exist (but I could criticize non-use or inadequate use once it did). For the subject of this discussion, whatever Dan achieved or couldn't achieve with CR 3.7 is irrelevant - my focus is to explore the merits of what Dan said about CR4 in respect of its "range setting" function. Readers may refer back to Annex 1 of my first paper where they will see that Dan finds the same "range setting" (photographers call this "contrast adjustment") issues with CR4 that he had with previous versions of CR. It is therefore completely reasonable for me to test this proposition using CR4 without regard to what Dan did or didn't do in CR3.7.

...... On to the second image, which *is* a fair example. It offers a red flower of the type that I showed in PP5E, and it limits the correction to methods available in CS2. Even though we have only the thumbnail PDF images to look at, it is clear from comparing them that Mark's final result is unacceptable,.....

Mark shows the original, uncorrected image in his Figure 21. Plainly, it is too
flat, since nobody has opened range yet. But doing so in ACR, as we all
should know by now, blows out critical detail in the lightest channel, in this
case the red. Mark shows and comments on this loss of detail in Figure 22. The color is brighter and more pleasing, but professionals would never accept it because of the detail loss.

First, in photography, we don't "open range" - we increase or decrease contrast. Second, increasing the contrast of an image in CR may, but does not necessarily blow out critical detail in the lightest channel. It depends on the image and the extent of the adjustment. This image was purposely driven to do so, in order to demonstrate that the detail is recoverable. Figure 22 was adjusted beyond its appropriate level of brightness and saturation in order to suppress detail for the purpose of demonstrating how an additional set of meta-data instructions brings it back, because it never got lost in the first place - being a raw file in a raw processor the original file data never gets damaged or destroyed.

In CS2, the only available countermeasure is a blunderbuss desaturation of all reds. That's what Mark does in Figure 23, his final version, .....One simply cannot do that to a picture that is supposed to portray vivid colors, as this flower should.

The image was shot in front of my house in Toronto, probably at least 600 miles from Dan at the time. Hence he didn't see this flower and has no accurate idea what the image should portray.

It puts the detail back into the lighter areas at the cost of muting the overall color. The final result is too gray to be acceptable.

The tools for de-saturating colours are more discriminating in CR4 relative to those of CR3.7, and I'm talking about what we can do to recover detail in CR4, not CR3.7. CR3.7 didn't have a tool for de-saturating reds in particular, so when speaking of CR3.7 in CS2 Dan must be referring to a general "blunderebuss" rather than a colour-specific one, but as I say, he's not talking about the tool I used. Figure 23 is a more accurate rendition of the flower than is Figure 22 - by design. If Dan had seen the flower and the raw file before issuing his critique, he would have been able to determine that the red in Figure 22 is both exaggerated and out of Ultrachrome K3/matte paper gamut, and therefore most likely out of any press gamut. The Figure 23 image is much more accurate and much more accurately printable. I verified this with Softproof and Gamut Warning in PS.

On to the first image, which is a still life of oranges and pink grapefruits in a
wicker basket. In addition to the question of loss of detail in the colored areas,
Mark notes that it arrives with a cast and suggests that it answers the
description of category #3 above. It does not. Oranges and pink grapefruits
are both light objects, as is the basket. Nothing of importance occurs in the
midrange. If a lime, a pear, an apple, and red grapes were added, then the
image would qualify.

Dan is changing the goalposts after the game has started. The specification of the image I made for my analysis conforms to the description of category (1) and partly category (3) which Dan proposed in message 18270 of July 26th, and I recited at the outset of my paper. It is about a predominant colour containing important detail as well as a cast - nothing more and nothing less. The image did not arrive with loss of detail in the coloured areas. I induced it as part of the testing. The basket is not all light - it is light and dark, but for purposes of this analysis the basket is irrelevant. As the curves and Histograms in the paper show, most of the important detail is within the mid-range to just below the quarter-tones. There was no question in Dan's original description of image category #1 about introducing a variety of colours, say by adding anything like pears, apples or red grapes to make it "qualify". And in any discussion with me about the selection of images, we can discuss the attributes of what "qualifies" before testing, or we can each select our own images and justify why they "qualify", but we won't have one party simply declaring the other party's images "unqualified". In this dialogue, there is a paradigm shift - everything is Q.E.D (Latin for "to be demonstrated").

Unlike the second image, where there was enough to compare between the
thumbnail images to know that the final result was unacceptable, here it
would be difficult to say how bad the image is. It's pretty hard to see detail in
yellows.

Anyone with adequate eyesight can see the presence or absence of detail sufficiently for the purposes of this exercise, and the predominant colours here are not yellow (hue angle 60 degrees), but a range around orange (hue angle in the range of about 30~40 degrees give or take a bit).

I would be very dubious that this is an acceptable result but without
seeing the actual file I could not say for sure.

All the more reason to ask for the file before trashing the conclusions reached with it. The beauty of using an image with the hues I used in this context is precisely the fact that detail is fragile, so if part of the argument is that increasing contrast with a composite curve over-saturates the colour and blows detail in bright predominant colours, this is a particularly appropriate and sensitive test image.

It is helpful, though, that Mark showed the progress of two channels as he
applied steeper and steeper S-shaped curves to the original. In the cherries
image of a few weeks ago, I explained how ACR uses a master-curve
approach to setting range. That is, it applies a single, uniform correction to
each channel, then averages them, performs other calculations, and
regenerates the final RGB document. This is a naive approach. Obviously,
ACR should contain channel curves as every professional-level program
does. But if there is only a single curve, then the correct approach is to
average the channels together before applying it, instead of applying the
same correction to the channels separately. Then, saturation could be added
in a variety of ways--including, if desired, as an automatic part of the curving
process.

The cherries image is irrelevant to this discussion because it is a JPEG. I'm discussing raw data files only, not pre-baked, pre-damaged JPEGs. In this context, Dan references his previous discussion with Jeff Schewe:

This has already come up once on this list, when Jeff Schewe was working on the cherries image that I had posted. Using only tools that were found in CS2, he was unable to get a satisfactory result. He then improved upon it by using Split Tone, a command found only in CS3.

Here is what Jeff responded by way of correction in ACTL message 18279 of July 27th:

Schewe: <To be accurate, Dan, I never tried opening your red sample image in Camera Raw 3 in Photoshop CS2...couldn't since the ability to even OPEN JPEG images wasn't added until Camera Raw 4 in CS3. I never opened your images in CS2-period. I think that was you, remember who opened them in CS2 by mistake?

And I did indeed "compensate for the detail loss" as noted by several other members of the list. No, it was not numerically a match but to my eyes, it WAS pretty close, particularly after you pointed out I lost saturation in the shadows and used split tone to bring them back.

And all this was done on a JPEG in sRGB mind you. . .not a raw image for which Camera Raw has been designed...

I still don't have access to any other raw examples where you claim CR does an unprofessional job...> End of Schewe quote.

Now let us turn to Dan's description of the math behind Curves in CR. The first essential point is that in CR, Curves alone do not set the contrast range. Much of this work should be done in the Basic Tab first. Hence there are symbiotic relationships between what one does to contrast and saturation in the Basic menu and what one does thereafter in the Tone Curves menu. If we're talking about the underlying math, we need to understand both and how they interact. It was already mentioned on ACTL that Thomas Knoll said much of this information is available in the DNG SDK, yet I see no reference in Dan's message indicating that he has studied the DNG SDK. Back in June Dan said he worked out the exact math underlying contrast adjustments in CR. A number of ACTL members, me included, have asked Dan several times since then to produce the spreadsheet, formulae and results he says he generated to fully explain the math behind CR Curves. Perhaps if this material were provided, it would contribute to a better understanding.

But as it stands, the description of the math provided in this message I find particularly obscure. What is a <<single uniform correction to each channel>> - is it an absolute amount, a constant proportion, and are there channels in CR to which these amounts are added? It goes on: <<then averages them>>: What gets averaged and why? What is the numerator and what is the denominator? Then it goes on: <<performs other calculations>>: what are these "other calculations"? Then it goes on: <<regenerates the final RGB document>> - well, it doesn't "regenerate a final RGB document". There is no such thing until after the image is rendered from CR into PS. But let us say it did regenerate something or other: what's the math underlying this process? It goes on: "This is a naïve approach". Dan is implying here that he knows more about what's naïve than the specialists in photography, digital imaging and digital imaging mathematics who have been programming and testing these functions with their clients for over two decades. Then it goes on <<Obviously ACR should contain channel curves>> - why "obviously" - what is so obvious about this? If it were so obvious why hasn't Thomas Knoll & Co. programmed them into CR long ago? Surely they know the issue and they know how to do the work, and they know what their competitors do. Has Dan tested OTHER raw processors which he says do have individual channel curves? Do they do a better job than CR4? Can he produce evidence to this effect? Then it goes on: <<The better approach is to average the channels together>>. Well, how is that done? - the formula that immediately comes to mind would produce a range of grey tones. Then somehow the image would need to be converted back to colour. So it isn't clear what's intended here. Enough said, all this discussion of the math is obscure and has no obvious operational significance relative to the empirical findings in the paper, so let's move onward.

..... In the red channel, which is the lightest, if an S curve is applied
as a master curve, contrast in the fruits will be reduced, but contrast in the
wicker, which would fall in the steep part of the curve, would be increased.
Similarly, the blue. Shadow contrast would decrease, but the wicker contrast
would increase.

As I said above, the basket is irrelevant. Not part of the Q.E.D.

The steeper the center of the S curve, the worse it is for the red channel. Mark correctly notes that nevertheless overall contrast does not appear to decrease as the curve gets more pronounced, but his analysis of why is not correct.
Overall contrast loss is avoided because the green channel, which is not
pictured, falls in the steep part of the master curve. As the green channel is
around twice as important as the red, and six times as important as the blue,
any improvement in it will cancel out or perhaps even outweigh any deficiencies in the other two channels. Nevertheless, there is no reason to
settle for a bad red channel (the blue is bad too, but unimportant). Therefore,
stressing again that I have not seen the actual image but only the thumbnails,
it seems unlikely that this is a satisfactory result. (The ACT course offers a
similar exercise with an image of a lime, so I'm very familiar with what type of
channel structure results in a good image--I've seen hundreds of people work
on it.)

All this is very interesting, but beside the point. Again, it would have been helpful had Dan examined the file before commenting on it. I can't find where in the paper I say that overall contrast does not decrease as the curve gets more pronounced. I wasn't discussing "overall contrast" - I was discussing the specific issue of what happens to saturation and image detail in bright predominant colours when contrast is increased with the CR Curve. So let's stay with the object of the exercise.

Needless to say I examined all three channels of the comparator images carefully before deciding which ones mattered most. Remember we are comparing changes of contrast, saturation and detail in the citrus from one state of the image to the next. Therefore what matters most about selecting channels is which ones change the most and in which ways. In this image, it was the red and the blue channels. Change to the green channel is much less pronounced between image states. This is not an image of limes which are predominantly green, but of oranges and pink grapefruits which are predominantly orange or orangey-reddish. The hue angle of limes is ~90 degrees. The hue angles of the important detail in my image are about 30 ~ 40 degrees. It doesn't matter how many hundreds of students have seen Dan's image of limes, this is not an image of limes. The important channels for this image are the ones I discussed for the reasons I illustrated them, and the explanation of what happens to the image from the Figure 5 state to the Figure 7 state is unambiguous.

Who said anyone is settling for "bad channels"? I'm not. Figure 5 is a successful photograph. Figure 7, by design, is not. Figure 9 is an improvement over Figure 7.

The inadequacy of this result could be demonstrated by comparing it to an
image properly corrected in Photoshop, which Mark purports to do--except
that he uses a master curve!

As I said in the paper, the only purpose of implementing the comparison in Photoshop was to show the comparative lack of impact on saturation by using a Luminosity RGB composite curve in PS, compared with the properly saturated results from using the CR curve. Again, because the CR curve is a one curve solution, it is compared with a composite curve in PS. At least I'm not comparing limes and oranges! The results - yet again - confirm the judgment of the Adobe programmers that images generally look more believable when contrast and saturation move in sync.

So, of course, it damages the image in the same way that ACR does.

It wasn't intended to damage or not damage the image. It was intended to show exactly what it showed, which isn't about damage in this case.

In something as orange as these objects, the red channel is light, the blue is dark, the green is in the middle. These are short ranges for the purpose of channel curves--easy to emphasize detail.>

The image contrast is sufficient for the purposes intended, can be and was varied sufficiently for the purposes intended and the impact on saturation and detail performed in the tests emerged as expected.

But for the purpose of a master curve, every tonal value from lightest to darkest is important in * some* channel. By definition, a master curve of any shape *must* damage at least one channel in an image of this type. Nobody with significant curving experience would work in Photoshop this way.

In a raw converter there is no damage to the original raw data period. Detail can be smothered and recovered as I have shown. This is always true in principle and people who know how to use the program can manage it in practice. The Annex is about working in CR, not in PS, so the comment is not relevant to the subject matter of the paper. If indeed there would necessarily be such damage in PS (but I am not convinced there always need be significant damage by definition), it is all the more reason to do as much of this work as possible in CR, where it is both easier and safe.

Certainly Mark is to be commended for having put these papers together and for suggesting ways in which people can work in ACR and discussing some of the alternatives within the modules are.

Thanks, much appreciated.

However, Because there is no valid comparison with other methods, and no way for readers to make the comparison themselves, the sum of what Mark's two papers demonstrate is that it is possible to apply certain corrections in Camera Raw that make an image look better than if they had not been applied. It is unclear that 40+ pages are required to make this point successfully.

Readers can make all the comparisons they want themselves. The images are available on request. There's nothing like people really trying all this stuff themselves to see what works better for them. Tens of thousands of people have seen these two papers and the feedback I'm getting suggests they are contributing much more than what is conceded here. In fact, they demonstrate that under a wide variety of imaging conditions, CR4 delivers professionally satisfactory results. But Dan need not believe me or my work. I'll simply refer him to Mikkel Aaland's new book on Lightroom, (which application has the same processing engine as CR4), where a dozen world-recognized professional photographers did a workshop in Iceland using nothing but Lightroom and came back with stunning, published results.

Now that the command has been improved in CS3, one could write a similar paper arguing that one should neither use ACR nor channel curves, but rather Brightness/Contrast.

This statement makes no sense to me. But whatever, "Brightness/Contrast" are legacy commands which most gurus advise people as a general rule to avoid. Nonetheless, sometimes a slight tweak of these sliders in CR's Basic Tab actually does produce a pleasing result - again, like always - depends on the image.

mages might be shown of elephants and the like,

Well, if we're talking about the usual grey elephants those are the kinds of images one would obtain using Dan's curves averaging technique which he recommended above. But of course there are also those White Elephants, such as "Brightness/Contrast" in Photoshop which may require a Selective Color Adjustment Layer and a slap on the display panel to excoriate. Cummon Dan, let's leave the elephants in the game parks.

In fact, using a combination of Brightness/Contrast on certain channels and Levels on others would likely get better results than Mark did on the first two images, because we would have the option of leaving some channel(s) alone if the commands would damage them.

"Brightness/Contrast" is not available on a per channel basis in either CR or PS, so I don't understand what is being proposed here.

I think my two papers have served the purpose of demonstrating the very considerable flexibility built into CR4.x for handling a wide range of raw image conditions. Let us all just acknowledge that fact and benefit from using it to the fullest of its potential.

Mark Segal
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Stephen Marsh
Fri Aug 17, 2007 3:47 am (PST)

André Dumas wrote:

Hello Stephen,

Your question: "Why was there no push for linear raw scanner sensor
data editing? " is an interesting one. I've been asking myself that
very same question.

Hi André, yes, it is interesting - but why indeed. Technical reasons? They both often capture light in a similar way (scanner & camera).

There was a movement to "capture down to the grain" in full high bit and optical resolution and density etc. The scanner software could be used without the original, to massage this data as if it was a raw scan on the glass.

SOOM from Scitex, Scan Once - Output Many (times) on the Oxygen software that drove high end units such as the Eversmart.

There is also HDR from Silverfast and perhaps VueScan also has a similar option?

But AFAIK, these scans/files are not linear, they are still gamma corrected.

Stephen Marsh.
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Stephen Marsh
Fri Aug 17, 2007 5:54 am (PST)

Chris Murphy wrote:

a.) It implies scanning film. Film is already rendered. It's
inherently not the same thing as a raw capture. A scan is a 2nd
capture of a scene. It's not the original capture. So scanners and
cameras are actually really different. They have a sensor technology
in common, but not much else.

Thanks Chris, the above is understood, but your last sentence seems to respond to what I was meaning about being similar...if a flatbed scanner's capture is linear by nature - just as a camera is, why not edit the data like we do with camera raw data (if both are raw sensor data in linear form). Why 'bake it' to a gamma space so early in the process?

b.) There is a single light source with a scanner and the primaries
used are based on film and light source. Scanners are ripe with
metamerism issues (when you start to scan things that aren't made
from film).

So, linear data and editing is not good with this data - even if one could access the raw linear CCD/CMOS data? What works well for a camera does not work well in scanning? I need more info here to understand your point.

If both capture data in a linear form, and there is great benefit for white balance, exposure and highlight recovery with linear raw sensor data editing - why was it never exploited before? I am really having a hard time grasping why one and not the other, when as you say in point (A) that the sensor technology is common, but little else.

c.) Handing over the scanner data stream to the end user isn't
something I've heard users ask for, let alone scanner vendors at the
time thinking their customers would have ever wanted. They prided
themselves on their proprietary software, which was rather
inextricably linked with the hardware, to process that stream and
allow it to be manipulated.

We are being told left, right and centre that linear data editing is the way to go for camera raw data, which is linear by nature. It is not simply good enough to demosaic and convert the raw sensor data to a high bit gamma corrected space for edits. So why are scanners so different? The capture is similar at the sensor level - so why convert to a gamma encoded space for post scan edits, when one could work in a linear space on the linear raw sensor data? Would one not gain better highlight recovery, exposure manipulation and white balancing?

As for Foveon and camera backs, those shots are considered rendered.
I'm not sure if the software for those cameras has direct access to
linear encoded, non-rendered data. It would then be considered "raw"
the one thing unneeded with them is demosaicing.

Interesting...To me, raw sensor data is just that - it is not qualified in my thinking as being from a Bayer, Foveon or scan back. Now I see that when most say CR, they mean the more common gray data that needs demosaicing and that other types of camera are considered rendered data and may not even offer raw sensor data, the benefit being that they are first generation and often high bit rather than second or third generation as with transmissive and reflective scans.

So, is changing white balance or exposure just like doing this with Photoshop and any other gamma corrected data? One does not get the benefit of the linear data in say highlight recovery? Surely if there is a benefit to working in linear mode with linear data as we do with CR files, then these cameras would be able to access it, even if they capture in full colour and do not interpolate the colour. This seems similar to my question on scanners not using a linear edit on sensor data workflow. If these cameras only major difference is no demosaic process, why do they not offer linear edits for white balance, exposure, highlight recovery etc?

I must be missing something simple/obvious here.

Stephen Marsh.
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: "Bob Smith"
]Fri Aug 17, 2007 7:13 am (PST)

It's been years since I've used it, but Silverfast had a linear option. The output could then be opened in Silverfast HDR and rendered into a more appropriate state for a particular use at a later time.

Bob Smith

Accurate Image • Bob Smith Photographer • Waco Texas USA
http://www.accurateimage.org
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: dave_cardinal
Fri Aug 17, 2007 8:37 am (PST)

It's been years since I've used it, but Silverfast had a linear
option. The output could then be opened in Silverfast HDR and
rendered into a more appropriate state for a particular use at a
later time.

Bob Smith

Bob--Yes, I was thinking the same thing when I read this thread. For many images I did all my correction in Silverfast/HDR and recall quite a bit of the correction being in linear space. But I have to admit that since I got my first D1 in 2000 I've been increasingly hooked on digital originals and the scanner sits idle.

That said, to amplify one of the points that I think Chris was trying to make, film normally already has a gamma curve, so even though we may have a linear rendering of the density of the film, we don't really have linear data of the scene itself with a scan (unlike a camera Raw file), so it makes sense that there should be differences in the best way to process each.

--David Cardinal
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Richard Wagner
Fri Aug 17, 2007 12:05 pm (PST)

On Aug 17, 2007, at 5:02 AM, Stephen Marsh wrote:

Thanks Chris, the above is understood, but your last sentence seems to
respond to what I was meaning about being similar...if a flatbed
scanner's capture is linear by nature - just as a camera is, why not
edit the data like we do with camera raw data (if both are raw sensor
data in linear form). Why 'bake it' to a gamma space so early in the
process?

To clarify the discussion, it is necessary to specify whether one is scanning negs or transparencies.

ISO 22028-1 (Photography and graphic technology— Extended colour encodings for digital image storage, manipulation and interchange—Architecture and requirements) is a good source of information on this topic.

Hardcopy scans and transparency scans are both considered "original-referred color encodings" and not "scene-referred color encodings," so they are fundamentally different from digital camera captures of scenes. (Output-referred and original-referred encodings are both sub-classifications of "picture-referred encodings.") Transparencies, from the scanning perspective, are not too different from scanning hardcopy, except that the color re-rendering transform that generates the output-referred color-encoding would typically need to make more significant changes since the dynamic range and intended viewing conditions for transparency materials are typically quite different than that of most other types of output media.

When scanning negs, there are two choices:
—the first going through a scene-referred image state;
—the second going directly to an output-referred image state.

To quote from the ISO:
"Due to the extremely wide dynamic range of colour negative film, it represents an excellent means for capturing scene-referred image data. To back out the characteristics of the film and scanner, a film unrendering transform is needed to infer the original scene colours from the film scan. In addition to compensating for the colour and tone reproduction characteristics of the film, it is usually necessary to account for the optical flare in the film camera and the colour response of the scanner. Also, it is usually necessary to apply a colour/density balancing step to account for the fact that colour negatives are typically captured over a wide range of exposure levels and using a wide variety of illuminants. Scanner compensation, color/density balancing and film compensation can be components of the film unrendering transform, or they can be be implemented as a series of individual transforms."

The following should be noted.
—The film unrendering workflow (going to a "scene-referred" image state) maintains the greatest amount of flexibility for repurposing and editing the image data. However, it also requires a more careful characterization of the film response in order to create the film unrendering transform.
—For cases where it is a requirement to match conventional photographic prints, the film rendering path (i.e., going directly to an "output-referred" image state) will generally be a more straightforward solution.

This ISO publication is a good read, if you haven't seen it. It also addresses the fundamental imaging pipeline used in digital still camera capture and processing.

--Rich Wagner
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Randall Hoffner
Fri Aug 17, 2007 2:49 pm (PST)

Don't know any of the particulars, but I note that Vuescan claims that they can output in the form of DNG files.

Randy Hoffner
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Stephen Marsh
Fri Aug 17, 2007 8:16 pm (PST)

Thanks to all who have posted so far.

The scanner stuff is not really important in this discussion and is a side issue, never intended to drown out my main question.

It would be really interesting to know more though on Foveon and scan back rendered data - if the only major difference is no demosaic process, yet there appears to be a rather different image editing workflow with no manipulation of raw sensor data.

I will have a look at the books mentioned, but my OP was not asking for a tome - just a simple diagram, bullet point list or paragraph list explanation, similar to the one in the links provided in the OP.

I would have thought that this was easy to do, for those that know so much about Adobe CR processing and display this knowledge in many posts to the list in recent threads on CR processing.

But nobody has been able to give me a link to a simple flow chart or diagram of the Adobe camera raw processing path - or even to write out the order of the steps that take place from the input of the raw grayscale data to the final rendered colour image image.

I suspect that it is a trade secret to some degree, in part if not full (as some aspects are likely noted in the DNG SDK etc). If my questions are not known, then please say so, you will not lose any respect for being human and not a font of knowledge. If you do know, then please list the steps or draw a little diagram or something! I am not above doing my own research and I will learn a lot in my search for these questions, but it would be helpful for both myself and I suspect other list members if we could have the basic steps listed on this site or follow a link to them.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested (reason for linear
Posted by: "Sabo, Lori"
Fri Aug 17, 2007 11:57 am (PST)

Andre,

I can think of one good case for wanting linear raw scanner sensor data editing -- digital scanning cameras or digital scanning backs for film cameras. (Like Betterlight). There the lighting/scanning conditions vary constantly. I would love to be able to adjust the linear output from my Betterlight directly in Camera Raw. We use it to shoot hi-res scans of artists' paintings for high-quality giclee productions, of course wanting the result to be as true to the original as possible. Some people also use them for studio still life or even landscape photography!

Lori Sabo
___________________________________________________________________________

Re: Range setting confusion
Posted by: André Dumas
Fri Aug 17, 2007 8:18 pm (PST)

Hello Dan,

In your message of 8/13/2007 you said:

I explained how ACR uses a master-curve approach to setting range.
That is, it applies a single, uniform correction to each channel,
then averages them, performs other calculations, and regenerates the
final RGB document. This is a naive approach. Obviously, ACR should
contain channel curves as every professional-level program does. But
if there is only a single curve, then the correct approach is to
average the channels together before applying it, instead of
applying the same correction to the channels separately.

When you say "performs other calculations" can you tell me what are those "other calculations" and what do they do? From my own experience I suspect that ACR's Tone Curve does not operate on the Raw data the same way that Photoshop Curves operates on the individual channels ?

I have been using Nikon Capture for the past 4 years (off and on, ACR is my preferred converter) and I have never found that I could achieve a better image because of the channel curves. Have you used Capture or another Raw converter and have you found that it produces better results ?

Personally I prefer using the Tone Curve of ACR or the Master Curve of Capture because adjustments are simpler and they maintain the color balance of the image, when I need to remove a cast I find that ACR's White Balance does that nicely, in Capture the same thing can be done with a channel curve but not if correcting the cast involves two channels, in such cases the Master Curve does a better job.

Does that make sense Dan ?

André Dumas
___________________________________________________________________________
 
Re: ACR Image Processing Pipeline Info Requested
Posted by: Mark Segal
Sat Aug 18, 2007 7:39 am (PST)

Stephen,

Maybe there is some misunderstanding about what you are asking for:

Are you asking for {A} the technical detail of what happens under the hood regarding the ordering of the processing steps in Camera Raw from importing the data, to demosaicing it to rendering it, or [B] the optimal order of using the image adjustment tools in the Camera Raw menus prior to rendering? Most of what I referred you and most of what we have been discussing on this List to deals with [B], about which we know quite a bit. Item [A] is of intellectual interest, but knowing more about it probably wouldn't contribute to improving image quality. There has been some recent discussion of [A] on Luminous-Landscape in several threads of the Digital Image Processing section in the Discussion Forum.

Mark Segal
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Dan Margulis
Sat Aug 18, 2007 10:36 pm (PST)

Stephen Marsh writes,

Scanners are not the same as digital cameras, but they are not far
removed either. The flatbed scanner at my day job has a very linear
tone response. Why was there no push for linear raw scanner sensor data
editing?

There was a market if there was a benefit (and perhaps even if not with good
marketing), before digicams scanners and scan software was a big issue, that
seems forgotten today with digicams being so common.

Traditionally this wasn't even a possibility. Until around the early 1990s scanners were analog devices. There was never a base file as such; the scanner offered the user an interface that in effect was gamma corrected (just as ACR does with its own curve commands). The curves and other commands were translated on the fly and applied directly to analog voltages (obviously linear), then retranslated into an output file that was written on the fly. This was not done for quality reasons but technical limitations.

When scanners became digital all the way, it also became possible to store pieces of the image in RAM. I don't know whether these temporary files were stored in gamma-corrected form but I doubt it, as it would have obsoleted past methods.

For years we have been warned about converting to or editing gamma
corrected RGB files in a linear gamma RGB editing space (with one
[in]famous exception endorsing this workflow). Then along comes camera
raw data, which is linear by nature and thus linear editing is
appropriate in this setting.

Linear editing is appropriate sometimes and not otherwise. As a rule, range commands and sharpening work better at a higher gamma and color and blurring are better handled at lower gamma, not necessarily linear. However, since unlike a scanner, there *is* a permanent file, and since the file arrives in a linear form, there's not a lot of point in changing it before it's necessary. In a raw module, one can apply overlays so that this correction looks and in fact is a low-gamma move but that correction appears to the viewer to be in high- gamma and the other correction seems to the viewer to have been done in CMYK.

Then André put his finger on the whole issue, saying, "It seems to me that having a 'raw scan', which would make repurposing possible, would be a good thing." And indeed it would. The problem is that we are so accustomed to thinking digitally that we does not occur to us that the film is itself the raw file. It is the record of what the camera captured. If detail is not present in the film, no scan can resurrect it, any more than a raw module can extract nonexistent data from a rile. If there is even a small amount, either a scan or a raw module can intensify it. The only limits as to what it can be repurposed for are what's in the file/film.

André's "raw scan", might not be just the film, but instead a very conservative scan that made certain that it did not blow out highlights or shadows or introduce any cast. Or, it could be my recommendation for Camera Raw--be conservative. Either way, it might not look good at first, but it would be easy to repurpose.

Dan Margulis
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Chris Murphy
Sat Aug 18, 2007 10:36 pm (PST)

On Aug 17, 2007, at 8:02 AM, Stephen Marsh wrote:

Thanks Chris, the above is understood, but your last sentence seems to
respond to what I was meaning about being similar...if a flatbed
scanner's capture is linear by nature - just as a camera is, why not
edit the data like we do with camera raw data (if both are raw sensor
data in linear form). Why 'bake it' to a gamma space so early in the
process?


Early scanners weren't digital, they used photo-multiplier tubes and had a circuitry designed to aid in encoding the already tone mapped, rendered image, into a limited bit depth. Very early scanners weren't scanning with as many bits because that was expensive. And linear encoding in 8bpc is a problem because we don't see that way. Linear encoding devotes a lot of bits to highlights and in 8bpc encoding not nearly enough are available for shadows. I think this tradition was simply carried on with scanners and their scanner software because the purpose was to eventually output the files.

Later on, it was possible to get some high-end scanners to scan and output a linear file, to which a scanner profile was applied so that it could be tone mapped (again through at least 16bpc precision) to whatever you wanted.

So, linear data and editing is not good with this data - even if one
could access the raw linear CCD/CMOS data? What works well for a
camera does not work well in scanning? I need more info here to
understand your point.

Linear data and editing is not a problem. For some kinds of blending it's preferred in that at least mathematically it's more accurate. But it does necessitate at least a 16bpc file. (We could get away with less bitdepth than 16bpc from the scanner itself.)

If both capture data in a linear form, and there is great benefit for
white balance, exposure and highlight recovery with linear raw sensor
data editing - why was it never exploited before?

White balance is a function to account for the spectral sensitivities of the RGB filter set in a device. Its behavior is rather different depending on a changing illuminant. There is no changing illuminant for a scanner so it's not relevant. This function for raw digital camera files is not a simple function for removing color casts. It recognizes both that the RGB filter set does not match the xyz cone response of human vision, and that the response varies under different illuminants.

As for exposure and highlight recovery, again those features work on raw files not merely because they're linear and we have 12-14bpc of data, but because those files are unrendered. If such extreme shadow or highlight data actually exists in the raw file, these tools can be used to visually expose the data that's there, via tone compression, that would otherwise have been clobbered by in-camera rendering to JPEG where it's simply gone. Film is analogous to JPEG in this respect (although the dynamic range is different) in that if it's not on the piece of film you can't get it. It's gone. If it's on the piece of film, but you haven't adjusted the scanner to set white and black properly, you also don't have the data. If the dynamic range of film exceeds the dynamic range of your scanner, then you may lose either or both extreme ends of detail.

Another big factor is that the mathematical functions related to tone mapping in (at least some) raw processors are new to digital imaging and have more capability than what existed in scanner software and even Photoshop.

Interesting...To me, raw sensor data is just that - it is not
qualified in my thinking as being from a Bayer, Foveon or scan back.

Well it does make a difference because if you're getting a tone mapped file it's likely rendered in some fashion to get there. If we were talking about 32bpc that wouldn't necessarily be the case because that's a lot of bits to the point where the tone reproduction curve really doesn't matter. But currently, a mosaiced raw file is clearly unrendered. But you can have demosaiced, linear data in a DNG and that too is considered unrendered. I'm currently just lacking information on what data Foveon based cameras we have access to so someone else would have to answer that question.

So, is changing white balance or exposure just like doing this with
Photoshop and any other gamma corrected data?

Completely different math is involved in the white balance slider in CR than anything you'd be doing in Photoshop. Could you find some elaborate sequence to arrive at an as good, or maybe even better result, given unlimited time? Sure it's possible. But it still would not involve the kinds of computations of these particular functions in CR. The camera's RGB filter set responds differently under incandescent light than it does daylight, and this is highly non- linear. CR has generic profiles for both of those illuminants in it for each supported camera make/model, and interpolates between those two points as you adjust the white balance slider. It's not merely correcting for a uniform color cast.

If you hand LR/CR a JPEG or a TIFF, there is no such thing as highlight recovery because there is no highlight data to recover. And you'll notice the white balance slider doesn't work the same at all, showing relative values rather than in Kelvin. What's in the file is already presented (tone mapped), and what's not there is simply not there. Whereas with a raw file, there may be be hundreds of levels of data above the default white point.

Therein lies what's going on with in-camera rendering, from which you get a JPEG directly out of the camera. Due to aggressive white point setting and contrast adjustment, there can be hundreds of levels of highlight data in the capture (in the internal raw file) that are simply discarded and never find their way into the JPEG. You can't recover what's not there. I must be missing something simple/obvious here.

New math, new tools, new concepts. The problems are essentially the same, but where things happen is very different. With film/scanner workflow, primary rendering occurs as a function of film chemistry. There is no UI, and that math was not available to the average person and considered highly proprietary. There was little interaction or ability to modify it. You either liked it or not. Much like in-camera JPEG today.

With digital, raw workflows, primary rendering is now an end user function. It has shifted from film manufacturers to the end user. This is a very significant event in photography. New tools, concepts, math, and now workflow, are now in demand to content with this. Previously it wasn't needed, or perceived to be needed, or possible really with the hardware and software we had.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Chris Murphy
Sat Aug 18, 2007 10:42 pm (PST)

On Aug 17, 2007, at 12:31 PM, Richard Wagner wrote:

To quote from the ISO:
"Due to the extremely wide dynamic range of colour negative film, it
represents an excellent means for capturing scene-referred image
data. To back out the characteristics of the film and scanner, a film
unrendering transform is needed to infer the original scene colours
from the film scan.

This is a rather industrious exercise and I have tried it before quite a while ago. Each batch of film can have rather different behavior in this respect, even when it's the same ISO, same conditions, same filters, same illuminant, etc. And then you have different software packages that deal with the orange mask and then inverting the image, which is proprietary.

But you could characterize the net behavior of the negative and a particular scanner (and its software), by shooting a suitable target properly on each roll of film, scanning it with particular scanner and particular scanner software (and a particular version of that software), and then make a profile for it. This way you could get back to some idea of scene colorimetry which is what is meant by a scene-referred image. But this is needless to say quite a project, and really as far as I can tell, impossible to do after the fact if you don't have targets shot on the film being used.

And given the dynamic range of color negative, I'd be concerned about the dynamic range of the scanner actually introducing problems in fussing around with getting back to scene colorimetry in the first place.

In any event, it's a tall order.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Richard Wagner
Sun Aug 19, 2007 10:30 am (PST)

On Aug 17, 2007, at 4:49 PM, Stephen Marsh wrote:

It would be really interesting to know more though on Foveon and scan
back rendered data - if the only major difference is no demosaic
process, yet there appears to be a rather different image editing
workflow with no manipulation of raw sensor data.

I don't quite understand what you're saying - "there appears to be a rather different image editing workflow with no manipulation of raw sensor data." Adobe has supported the Foveon RAW file since version 2-point-something. If you want to see how to handle Foveon RAW data, you simply have to read the code in Dave Coffin's DCRAW source file or contact him - he personally deciphered the metadata in the Foveon RAW file. (http://cybercom.net/~dcoffin/dcraw/ ) There *has* to be manipulation of the raw sensor data, or you never get a rendered image! The Foveon sensor is fundamentally different with its stacked pixel design, and many of the early algorithms will necessarily be different because of that design, but there is a lot in common with other RAW files. The entire process is fundamentally the same.

But nobody has been able to give me a link to a simple flow chart or
diagram of the Adobe camera raw processing path - or even to write out
the order of the steps that take place from the input of the raw
grayscale data to the final rendered colour image image.

Well, AFAIK, Adobe has not published such a flowchart, nor would I expect them to. I'm sure Apple, Nikon, Iridient, Bibble and others would love to have all the details - even in block diagrams - which they haven't published, either. In addition, Adobe recently revamped their RAW processing pipeline in CR Version 4. Perhaps Jeff Schewe's book (due out in October) will give more specifics that have come directly from Adobe. I don't have that kind of access, although Thomas Knoll and other engineers at Adobe are very good at answering relevant, specific questions about CR, and they are downright lightning fast at responding to reported bugs or deficiencies in the CR or LR processing pipeline. The paper by Karl Lang that was recently cited here also described the RAW process well, if not all of the technical details.

If you're willing to settle for a description of how the *old* Camera RAW worked, there's a chapter from the *old* Camera RAW book by Bruce Fraser available online for download. It is still a good read - and it's free.

http: //www.macworld.com/2005/01/downloads/cameraraw.pdf

Bruce also penned this:
http: //www.adobe.com/products/photoshop/pdfs/understanding_digitalrawcapture.pdf

I've posted two images in a folder called Imaging Pipeline on the ACTL site that show DSC processing pipelines similar to what you've requested. I've added additional info to the ISO file in red. The articles that the flow diagrams came from are very good. Usually, though, what you find in articles are the specific processes - e.g., demosaicing, but you probably don't want that much specific detail. How developers put the pieces together is proprietary - as well as many of the specifics of the processes themselves.

I hope this is what you were looking for.

--Rich Wagner
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Chris Murphy
Sun Aug 19, 2007 10:50 pm (PST)

On Aug 18, 2007, at 7:56 PM, Chris Murphy wrote:

White balance is a function to account for the spectral sensitivities
of the RGB filter set in a device. Its behavior is rather different
depending on a changing illuminant. There is no changing illuminant
for a scanner so it's not relevant. This function for raw digital
camera files is not a simple function for removing color casts. It
recognizes both that the RGB filter set does not match the xyz cone
response of human vision, and that the response varies under
different illuminants.

OK, clean up time. That's not quite right. The camera's spectral sensitivities are what they are, they don't change with different illuminants. But the job of a color matrix (or lookup) is to account for the difference in the RGB response of the camera and human vision. As the illuminant changes, that difference is non-linear. That is, it's not just a simple color cast.

The camera's RGB filter set responds differently under
incandescent light than it does daylight, and this is highly non-
linear.

Again, the RGB filter set actually responds the same regardless of illuminant but we want to to respond differently, because human vision does. There is chromatic adaptation occurring, and also the colorimetry of objects in the scene will vary depending on the illuminant. Therefore the color matrix (or lookup) needs to handle variable illuminants, which is what the white balance slider in CR does.

Chris Murphy
Color Remedies (TM)
New York, NY
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Stephen Marsh
Mon Aug 20, 2007 1:51 am (PST)

Mark Segal wrote:

Stephen,

Maybe there is some misunderstanding about what you are asking for:
Are you asking for {A} the technical detail of what happens under the
hood regarding the ordering of the processing steps in Camera Raw from
importing the data, to demosaicing it to rendering it, or [B] the
optimal order of using the image adjustment tools in the Camera Raw
menus prior to rendering?

I don't know how I could have made it clearer, I stated in the OP:

"I am not talking of what order the user may move the sliders, but the order that is executed by the rendering engine". I supplied links to generic CR examples and asked for similar in specific regard to ACR.

But perhaps with my many musings and questions, the post was not as clear as I thought it was.

Thanks anyway Mark, I think I have some answers now (I looked at the LL forums but the threads did not leap out at me and I had no time to dig deeper).

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Stephen Marsh
Mon Aug 20, 2007 1:53 am (PST)

Thanks to all who responded to clear up my thinking on scanner CCD vs DC CCD capture, including Chris & Dan. I was getting caught up with the capture device recording light in a similar way - but not taking the next step and factoring in what the light being captured actually *is* (what it represents, captured).

Thanks for your patience!

Richard Wagner wrote:

I don't quite understand what you're saying - "there appears to be a
rather different image editing workflow with no manipulation of raw
sensor data." Adobe has supported the Foveon RAW file since version
2-point-something. If you want to see how to handle Foveon RAW data,
you simply have to read the code in Dave Coffin's DCRAW source file
or contact him - he personally deciphered the metadata in the Foveon
RAW file.

That was in response to an earlier reply in the thread that stated that Foveon and scan back output was likely considered rendered data (not CR) and that a CR file was not known to be a part of the user workflow.

Your post indicates otherwise in part, that Foveon are not considered rendered files and that there is a CR file available where one can take advantage of white balance, exposure and highlight restoration on linear raw data.

Can you also confirm that scanbacks do not deliver a CR file and their output is a gamma corrected TIFF or other rendered file?

There *has* to be
manipulation of the raw sensor data, or you never get a rendered
image!

Yes, indeed Rich - but just as users who shoot rendered JPEG files in consumer models do not have access to the CR file, it does not mean that the camera does not shoot a CR file before it bakes the JPEG/TIFF and discards the CR internally, all without user intervention. That statement was in regards to a workflow where the camera does not offer the photographer CR data and only processed/rendered output.

Thank you for the requested basic illustrations, it does help me - and I hope others to understand what is basically taking place.

Stephen Marsh.
___________________________________________________________________________

Re: Do Your Curves Throw You a Curve?
Posted by: Stephen Marsh
Mon Aug 20, 2007 8:20 am (PST)

I have uploaded two sRGB renderings of a NEF file to a directory in the Photo section of this list, one with Nikon SW and the other with ACR, using as shot WB with no exposure compensation with no boost in the saturation slider etc.

http: //tech.ph.groups.yahoo.com/group/colortheory/photos/ [Nikon & ACR Detail & Saturation]

This CR image makes a good test image for various image editing tasks (camera raw rendering, CR tonal edit tests, profile conversions, sharpening etc) as it contains detail in areas that are sensitive to detail loss with raised saturation.

I once mentioned this image in an older post unrelated to this topic:
http: //tech.groups.yahoo.com/group/colortheory/message/15607

The original CR file can be found here (3.5mb, click the thumbnail pic):
http: //www.outbackphoto.com/workshop/NEF_conversion/nefconversion.html

Regards,

Stephen Marsh.
___________________________________________________________________________

Re: Range setting confusion
Posted by: Dan Margulis
Mon Aug 20, 2007 8:24 am (PST)

André writes,

When you say "performs other calculations" can you tell me what are
those "other calculations" and what do they do?

They are attempts to improve fidelity in certain color ranges. Unless you are doing something unusual, such as applying a strong inverted S curve, the likely noticeable areas would be that light cyan-blues, such as found in many skies, become darker and shift toward blues, and dark orangy-reds become lighter and shift toward yellows.

It has been implied, although I'm not sure it has been stated outright, that there is also a supplementary move to increase saturation in lighter colors. This is not correct. The increase in saturation results from the application of the master curve.

From my own experience I suspect that ACR's Tone Curve does not operate on the
Raw data the same way that Photoshop Curves operates on the
individual channels ?

Discounting the above-mentioned adjustments, applying the ACR curve yields very, very similar results to applying the same curve to the composite RGB. Much closer than, say, lightening the midtone in Curves vs. lightening it the same amount in Levels.

If you attempt to demonstrate this yourself, it won't work as your system is currently configured. This should not bother you, as I can think of no useful reason for knowing that the curves are nearly identical as opposed to merely being similar. However, I have suggested to Mark Segal that he should invest the time to see what steps are needed to make Photoshop emulate ACR behavior closely, and to explain them to anyone who is curious.

I have been using Nikon Capture for the past 4 years (off and on, ACR
is my preferred converter) and I have never found that I could
achieve a better image because of the channel curves. Have you used
Capture or another Raw converter and have you found that it produces
better results ?

Any method, Photoshop or otherwise, that uses channel curves should produce better results on certain images. The deficiencies of master curves have been well understood for a very long time. However, most of the time the damage is inconsequential. Either the moves being made are too minor to make a difference, or the image doesn't fall into one of the categories that have been described in this thread, where the damage to detail has been demonstrated and discussed several times.

Personally I prefer using the Tone Curve of ACR or the Master Curve
of Capture because adjustments are simpler and they maintain the
color balance of the image, when I need to remove a cast I find that
ACR's White Balance does that nicely, in Capture the same thing can
be done with a channel curve but not if correcting the cast involves
two channels, in such cases the Master Curve does a better job.
Does that make sense Dan ?

I do not understand the last three lines, but I agree that the rest is usually workable, especially if time is short. I would keep an eye out for images that fall into the three categories I desribed to Mark. I would also acknowledge that while a master curve may not damage contrast much, it can't help it as much as channel curves can. So I would accept that supplementary curves might be needed in Photoshop.

Master curves are a lot like Levels, or Brightness/Contrast: a poor idea in principle, but they work acceptably a lot of the time, provided you're careful. The problem with ACR is that it insists that we use them *all* of the time, which is unworkable for a quality-conscious user.

As I will no longer be participating in this thread, and will be participating in the list only on a limited basis for the next couple of months, let me close with a reference to some of the palaver that has occupied this thread about terms and what they mean, and as to what was generally known and when.

The Makeready Archive at ledet.com/margulis includes an expanded form of the 1996 article "Defanging the Curves Vampire". The version posted is as it appeared in my book "Makeready", which appeared later in 1996.
http: //www.ledet.com/margulis/makeready/MA21-Defanging.pdf

The column is aimed at beginners, because in 1996 curves were not nearly as widely used as they are today. This is perhaps my most permissioned column. It has been translated into several languages and used as part of the curriculum in at least half a dozen colleges.

The word "range", or variants, used as I have been using them throughout the thread and as professionals have been using them since I entered the industry, appears 15 times. Not only does "master curve" appear, but there is an entire section headlined "Beware the Master Curve!" explaining to beginners why they should avoid the practice. There is also a glossary section, where neither word appears. From which, I would deduce:

a) In 1996, the terms "range" and "master curve" were so firmly entrenched that the leading author in the color correction field could use them freely without further definition, and not even bother to include them in a glossary.

b) That the range-opening routines of 2007 ACR might reasonably be considered crude by the standards of what a beginner should have known in 1996.

Happy curving, and best wishes.

Dan Margulis
___________________________________________________________________________

Re: Range setting confusion
Posted by: Dan Margulis
Mon Aug 20, 2007 11:23 am (PST)

I intend this to be my final contribution to this thread, for reasons that should be evident in the second of two parts.

Mark Segal writes,

I never said my raw images are not available. Interestingly, Dan did not ask for
them before he issued his critique of my paper or his comment that my files
are not available;

I was never given a realistic opportunity to make such a comment. This list exists because far more people wish to correspond with me personally than I can accommodate. I have many other demands on my time than this list and responding to unsolicited e-mail. As pointed out in my recent newsletter, and in both of my books, the average time to get a personal response from me to a technical query is typically four to six weeks.

Mark has indeed sent me his papers in advance, offline. They raise many technical points. They require me to read 20-page documents. He has given me, as I recall, 48 hours in one case and 72 in another, to respond before actually posting the papers. I am sorry, but I cannot possibly answer private correspondence of this complexity in so short a time. As a courtesy to Mark, who has been a valued member of the list and whom I consider a friend, I have given my responses higher priority than other matters that I considered more important. But I cannot respond in the time frame he wants. Discussion of this new paper began on the list before I could respond privately to Mark, and at that point messages to him saying how the paper should have been written were more properly directed here.

On to the question of file availability. Mark has kindly offered to make his images available to anyone who e-mails him. OTOH, a list member recently wrote,

"If ...the raw images cited are not available for download, the citation is invalid
and the claims predicated on those images are invalid. Your claims have not
been cited with downloadable raw files that demonstrate the claims...It's
intellectually dishonest, and anti-academic to make claims about any raw
processor, while not providing raw files, for download, that demonstrate those
claims."

Personally, I think that the last sentence is over the top, particularly since the way I interpret the words, it brands both Karl Lang and the late Bruce Fraser as intellectually dishonest and anti-academic, as well as Mark. And I would say that "download" might be construed to include any method of getting a large volume of copies into the hands of whoever wanted them, without the claimant having knowledge of who might be working on them. But I would say that "available upon request by email from claimant" is inadequate, for three reasons.

First, when one is reading a book or article and feels that it is necessary to consult the images to see whether the author knows WTF he's talking about, my experience suggests that one wants those images right then and there, and does not wish to wait however many days it takes to try to contact the author by email and explain what he is thinking of doing with the images, etc.

Second, while noting Mark's opinion to the contrary, these PDF images are too small to make intelligent judgments about quality. I can see enough detail loss and color issues in the second image to state categorically that it is not of acceptable quality, but I don't know whether it's just marginally bad or a genuine stinkeroso. While I do suspect that the first one is also poor, I am not able to state this absolutely based on the PDF. The third image, which I did not comment on because it uses a CS3-only command, is an more compelling example. Mark's strategy there was to open range, which given the master-curve structure automatically blows out detail, and then to attempt to reduce the damage by desaturating, and then attempt to reverse part of the desaturation damage by using the Vibrance command, which itself has a heavy saturation component. In short, we have not just a counter-correction here but a counter-counter-correction, and not just one, but two saturation moves that affect contrast. As we discussed recently in the B&W command thread, such saturation moves are very prone to enhance noise and artifacting. So are counter-corrections. If we were seriously evaluating this image, we'd have to be alive to the possibility that Mark's technique had provoked such a problem. If it did, we wouldn't be able to see it in the PDF.

Third, putting out images gives the author a certain credibility because everyone realizes that it is a dangerous thing to do. Every time I show an image and make the original available, I run the risk that somebody who knows more about color correction than I do may decide to make a fool of me by showing people a much better version. That used to happen every now and then, and it was highly embarrassing. I've been luckier recently, but I'm not sure that Mark would be. I think that posting the raws that went with the first two images of this second paper might so irritate some professional retoucher (not me, I hasten to say) into deciding to teach a lesson to anyone who dares to defend using master curves as a serious color-correction method.

For these reasons, it was likely a wise decision for Mark to avoid posting them, and also to decline my invitation to test his color-to-BW skills against the known results of skilled folk.

With regard to the images themselves I would say that we have reached the point of being able to agree to disagree. David Marley put it best a few weeks ago. It is all a question of what constitutes acceptable quality. Many people use master curves, or Levels, or Brightness/Contrast, and are happy with the results. Mark says he is happy with his, which is fine. I agree with him that many people also would be happy with them, and that many photographers are happy with the results they get from making extensive use of ACR also.

However, I would remind Mark that the title of my book is *Professional* Photoshop. As its first edition appeared when few professional photographers even knew what Photoshop was, let alone how to use it, the "Professional" refers to professional retouchers, and to those who aspire to that type of skill set. It is an honor to have on this list a dozen or more photographers who have gotten to that level through hard work and dedication. Nevertheless, acknowledging that such exceptional photographers exist, it is still true that there is a sizable gap between the skills of photographers *as a group* and retouchers *as a group*. By the standards that I teach to, the first two images that Mark showed in this second paper would be unlikely to pass muster.

RE: THE FIRST IMAGE (Citrus Still Life)
Dan wrote:
On to the first image, which is a still life of oranges and pink grapefruits
in a wicker basket. In addition to the question of loss of detail in the colored
areas, Mark notes that it arrives with a cast and suggests that it answers the
description of category #3 above. It does not. Oranges and pink grapefruits
are both light objects, as is the basket. Nothing of importance occurs in the
midrange. If a lime, a pear, an apple, and red grapes were added, then the
image would qualify.

Mark replied:
Dan is changing the goalposts after the game has started. The specification of
the image I made for my analysis conforms to the description of category (1)
and partly category (3) which Dan proposed in message 18270 of July 26th, and I
recited at the outset of my paper.

Saying that it is partly category 3 is like saying that a woman is partly pregnant. Category 1 specifies strongly colored objects containing important detail. Category 3 specifies a uniform cast AND important objects in many ranges. Without both elements, the picture doesn't qualify, partly or otherwise. If saying that a picture with a uniform cast alone partly conforms to Category 3, then surely it must be valid to say that a black and white image containing important detail partly conforms to Category 1.

RE: THE SECOND IMAGE (Red Petunia)
I said that the flower had turned out too gray to be acceptable, and Mark replied,

The image was shot in front of my house in Toronto, probably at least 600
miles from Dan at the time. Hence he didn't see this flower and has no accurate
idea what the image should portray.

R.W.G. Hunt observed, "It is, however, a fact that the average person generally feels competent to appraise the colours in pictures taken by persons other than himself, of objects which he has never seen, at times when he was not present." In my own experience with advertising work, in a good 90 percent of cases the person making the final decision on the color fell into the above category.

When working in ACR/CS2 and confronted with a picture of a flower as red as this one or the similar image in PP5E (and, presumably, the raw file posted by Ric Cohn), we have our choice of which of three defects to accept: 1) a flat image; 2) a colorful flower with detail blown out; 3) a dull flower with detail retained. Mark chose option #3 and attempts to defend the way it looks. I would point out that even if the objective is a dull flower there are technical issues that would preclude acceptance of Mark's image, but the bigger point is that the client may not agree with Mark's aesthetic judgment, and may demand something more strongly colored, yet fully detailed. In that case, in ACR/CS2, Mark would be SOL.

RE: THE THIRD IMAGE (More Red Flowers)
I refused to comment on this image, stating that it was irrelevant to any criticism of my writing, since it depends upon a command found only in CS3, whereas everything I have written about is in a CS2 context. Mark replied,

As I've told Dan well before he posted his response to Annex 3, my papers
are forward-looking, hence he knew that before he questioned my context. So
why is he doing so now? I also told Dan at the same time that I'm not criticizing his book, he won't find any criticism of his book in my papers, and I would not be
silly enough to criticize not using something that didn't exist (but I could
criticize non-use or inadequate use once it did).

I confirm that Mark told me this offline. The problem is that he didn't tell any of the thousands of readers claimed to have read his paper. My name pervades both papers right from the start. These papers are being touted on the web as being a refutation of what I have written. Surely it would have been proper to include a disclaimer on this image to indicate that it used a technique unavailable to me when I wrote. This is particularly true when the stated objective of the paper is to compare an ACR workflow with a workflow that is notoriously associated with me--and then is never discussed again.

RE: THE PURPOSE OF THE PAPER
I criticized both papers because the stated purpose was to compare working extensively in ACR against opening conservatively and applying channel curves in Photoshop, whereas the only mention Mark made of a Photoshop alternative involved the use of master curves. He replies.

As I said in the paper, the only purpose of implementing the comparison in
Photoshop was to show the comparative lack of impact on saturation by using
a Luminosity RGB composite curve in PS, compared with the properly
saturated results from using the CR curve. Again, because the CR curve is a
one curve solution, it is compared with a composite curve in PS.

Yes. That's what it said LATER in the paper. But the paper OPENS as follows:

"The purpose of this essay is to explore whether one achieves superior image quality by [A] maximizing the use of White Balance and luminosity (contrast and brightness) adjustments in Adobe Photoshop Camera Raw (CR), followed by further adjusting luminosity of the rendered image in Photoshop (PS) using the RGB composite Curve (PS-RGBc), or instead [B] rendering the raw image with little adjustment in CR and using the PS individual channel curves to adjust White Balance and luminosity.

In my opinion "individual channel curves" and "Luminosity RGB composite curve" do not mean at all the same thing. It is my experience that when technical papers state in the first sentence that they propose to explore something, there is ordinarily some effort, however lame, to do so at a later point in the paper. Here, there is none. I would be happy to confirm or refute any statements in the paper as to how individual channel curves work, but there's nothing there at all. The whole point of my recommendation to avoid major moves in ACR is that master curves are a bad idea. What master curves do in Photoshop, in Luminosity mode or otherwise, do not address anything suggested in the opening paragraph.

And speaking of master curves, it is now time to end this and go on to a somewhat longer second half, which has been retitled "Some Hypothetical Images."

Dan Margulis
___________________________________________________________________________

Re: ACR Image Processing Pipeline Info Requested
Posted by: Mark Segal
Mon Aug 20, 2007 8:26 am (PST)

Stephen,

Now that you resurrect that quote from your OP it is indeed clear what you were seeking - I must have missed it.

Sorry for the confusion.

Mark Segal
___________________________________________________________________________

Re: Some Hypothetical Images (was: Range-Setting Confusion)
Posted by: Dan Margulis
Mon Aug 20, 2007 12:40 pm (PST)

This is the second half of my response to Mark Segal, who writes

Back in June Dan said he worked out the exact math underlying contrast
adjustments in CR. A number of ACTL members, me
included, have asked Dan several times since then to produce the
spreadsheet, formulae and results he says he generated to fully explain the
math behind CR Curves. Perhaps if this material were provided, it would contribute to a better understanding.

That's not quite accurate. I said that ACR used what amounted to a master curve in its range-setting routines, and this produced an increase in saturation as the image got lighter. This produced the sort of reasoned, factual, and dispassionate discussion typical of my cheering section on this list, such as,

1) "This is pure, unsubstantiated, meaningless horse manure...To state that * any* RAW processor is equivalent to using "master curves" in Photoshop is sheer nonsense."

2) "You keep saying this kind of unsupported nonsense...This idea of a master curve is absolutely not how the product handles the raw data, not even close. You really need to have a conversation with Thomas Knoll, Mark Hamburg (if they'd even talk to you) or read Bruce Fraser's Real World Camera Raw. You are either ill informed about the processing pipeline or you're making this stuff up!"

3) "Facts have been presented that directly contradict your supposition, and you keep saying the same thing. You have used a single image and a single metric to base your claim that it "uses a master-curve approach", that being the saturation change that occurs. There are other behavioral differences between curves in ACR and Photoshop that are considerable, and make them significantly different, and makes your characterization that CR tone curves works in the way you describe factually incorrect. The exact mathematics of this function are in the DNG SDK. It does
not jive with your description, at all"

As I am not a believer in troll-feeding, these all were diverted to my killfile, but there were a couple of similar comments from people whose phraseology was more elegant. Jeff Schewe, who many consider the leading living authority on Camera Raw's functionality other than full-time Adobe employees, in discussing the ACR tone curve, stated flatly, "Camera Raw doesn't use a Master Channel adjustment like Photoshop."

Andrew Webb also stated firmly that I was wrong in saying that a master curve was in play. I replied that I was right, and explained how to look for the telltale evidence. He wanted to know how I happened to acquire such knowledge. I replied with a lengthy description of how I tested ACR, which is about the same way I test any other feature or other imaging product.

If anybody was seriously interested in learning how the module worked rather than in trolling, there would only be three possible excuses for not knowing that the master curve is there: 1) being too lazy to run the tests; 2) being too ignorant to know how to do it; 3) being so credulous that they believe whatever they read in a book or whatever a programmer tells them.

My cheering section, falling evidently into one of the categories listed above, launched a campaign to see how many messages they could post demanding my notes. My friends, and especially my wife, would have been able to tell them that my work notes are approximately as useful to anybody else as would be a text written in Sanskrit. Nevertheless, since the thread began, this list has been deluged with 24, count 'em, 24, posts that include the word "spreadsheet". This mindless hassle-Dan-for-the-sake-of-hassling-Dan behavior is to be expected from most of the people who posted them, but I do expect more from Mark and am disappointed that he would bring this matter up.

The claim is that ACR uses a master curve when it sets range, thus adversely affecting detail and color balance in certain images. Having made that claim, I am responsible for offering proof, which I have done both here and in my book, properly peer-reviewed. There is no theory I have ever heard of in academia that would require me to offer private notes. Whether I know exactly how ACR functions matters not a bean, as long as I can show that it uses a master curve. If I do happen to know, exactly how I *came* to know is of zero significance. It is therefore regrettable that this thread operates under a civility warning, as otherwise I could offer a suggestion as to what might be done with demands for a spreadsheet or other private testing notes with respect to ACR or any other Photoshop function.

This thread has gone on far too long, and I take my leave of it. Before doing so, since Mark feels that a PDF is sufficient to show image quality, I have uploaded one to
http: //www.ledet.com/margulis/misc/ACR_Master_curve.pdf

As noted by David Marley about a month ago, master-curve behavior pervades contrast commands in ACR. But since the tone curve is the command that Jeff has specifically denied uses a master-curve format like that of Photoshop, that's the one being shown here.

The file shows variants of three raw images of very different subjects, one image per page. I am not the photographer, but by coincidence I happen to have a permission to use these images in any way I like.

Since the purpose is to test one specific command to see if there is a master curve present, the top left image on each page is exported directly from ACR with everything zeroed out. The bottom left image is created in ACR, using the tone curve command only. It is not supposed to represent my idea of good color. It is supposed to represent what happens when we apply a vigorous curve of any one of the three major types: linear (note the blown-out highlights on the first page), adjustment of a single intermediate point (on page two, lightening of a single point slightly lighter than the midtone), and S Curve.

The bottom right image (which is also always enlarged and shown next to a similarly enlarged version of the bottom left one) represents a different approach, one that starts with the top left, completely unadjusted, version and continues in Photoshop. Each one is run through an identical Action to try to emulate ACR behavior. The most important part of the Action, though not the only one, is the application of a master curve to the RGB file. And not just any master curve--the specific, exact curve applied to produce the bottom left version.

Now, these bottom right images are hypothetical. How could they be otherwise? We have repeatedly been informed that the internal workings of this tone curve are complex, that they involve specialized proprietary algorithms, whose behavior cannot be derived by outsiders nor emulated by anything in Photoshop. Plus, these are pretty strong master curves. Unless ACR uses not just any old master curve but one that functions almost identically to Photoshop's, we couldn't get remotely near to the resemblance we see here. Yet Jeff tells us that the actual operation is nothing like a Photoshop master curve. And we are also advised that it's "not even close", that it's "meaningless horse manure", and that it any such similarity is ruled out by Adobe's documentation.

If I actually had the files and action that would disprove these comments as devastatingly as these hypothetical images do, I would deny it. Why put the list through more months of hell? There would be 24 demands each for more images, for another set showing Brightness, and Fill Light, demands for my notes, demands for history logs, for my late mother's birth certificate, and so forth. Plus, what would be gained? For those who truly are interested in knowing how the command works, it's easy enough to experiment. For those only interested in the politics of personal destruction, being immune to concepts such as logic or demonstrations, it wouldn't make any difference anyway. And to the people who actually care about quality results, the issue is that ACR only possesses a master curve, not whether it is possible as a parlor trick to show Photoshop moves that accurately emulate it.

Imagine the ramifications, though, if such files existed. It would mean that this command, which is being represented as being something hugely modern, complex and sophisticated is instead at its heart nothing but a common, down-and-dirty, fleabitten, buttugly braindead worthless bleeping master curve that could have been lifted directly from Photoshop 2 and which sophisticated Photoshop 2 users would have known not to use, with a couple of flowers and ribbons attached to it with chewing gum as disguise. It would mean that many of the people who have hyped ACR the hardest actually have only the vaguest conception of how the module really works. If these images exist, then how can anybody take *anything* they said about the program seriously? I've seen people make fools of themselves in this industry before, and have done it myself upon occasion, but I don't recall anything quite as embarrassing as the position some of these people are in, if such files exist.

I imagine that any such files would withstand scrutiny in their high-res forms. I'd think the channels in each document would be nearly indistinguishable visually from one another. The two reds in the first image, I'd guess, would be the closest, and might have a mean pixel-to-pixel variation of on the order of .96 levels. Then again, if I were really making such an action for purposes of discussion, I'd probably only be taking 15 minutes or so and there would probably be some rough edges. For example, it might have some questionable renditions of reds, as in the third image. I'd guess that a better Action would probably get considerably closer, but I don't see the merit in taking the extra time.

In Las Vegas in a couple of weeks for Photoshop World, I expect to place some relatively large bets. While I will not confirm that these images currently exist, I can confirm that I would accept some even larger bets that I could quickly arrange to produce them.

This would be rather a waste of people's money. A much better bet, and one that I would be prepared to make, is that Mark Segal would be able to come tolerably close to the results shown in this PDF. and I think there are some compelling reasons for him to try. If he does, it's not a beginner-level exercise, but it's not all that hard either. It's easier now to investigate what's going on in CS3 than it was when I was researching CS2. Plus, some useful hints on how to do it lurk in my response to André's post today.

Mark, if I were trying to promote the use of ACR for use as something more than a means of opening images simply, I'd start by asking whether the purpose was to produce images that were good enough for most people, or whether it also is aimed at people like me and those of my skill set, of whom many monitor this list. If the latter, then you must be aware that we are skeptical of products claimed to be as revolutionary as this one, as we have each encountered around 1,500 other products making similar claims. You must be aware that we are interested in how a product actually works, and not in how the vendor believes that it works. We have a well-developed ability to differentiate a critical element from an attractive bauble. And we have a highly refined ability to recognize when the people hyping a product don't know what they're talking about.

Under these circumstances, the very last thing that's needed is a demonstration that the leading spokespeople for the product have a fundamental misunderstanding of the way the product's most critical feature behaves.

If you expect your message to be heard iby such people, you need to establish that you are capable of speaking on their terms and also that you have a mind of your own that operates independently of the received conventional wisdom about the product.

If you figure out a way to emulate the behavior of the ACR tone curve within Photoshop, as I did, then your credibility is greatly enhanced. To rely on my statements and demonstrations about how the command works is as fatuous as to rely on Adobe's. André, for one, wants to know the specifics of how this command works, and there must be others. I am not going to tell him. It has no bearing on what I teach and I have no interest in getting into a pie fight. If you, as a neutral observer, are willing to step in and say look, I don't care what Adobe says and I don't care what Jeff says or what Dan says, I've done the investigation myself and here is what I've found out--that would be a great service not just to your cause but to yourself.

I thank both you and Jeff Schewe for the constructive contributions to this thread, from which I now retire.

Dan Margulis
___________________________________________________________________________

Re: Some Hypothetical Images (was: Range-Setting Confusion)
Posted by: "Andrew S. Webb"
Mon Aug 20, 2007 2:49 pm (PST)

I stated that because that is what I had been told by people who are in a position to know. Turns out that they hadn't checked thier facts sufficiently, and only half-understood the situation. As I did my own tests, I found that Dan is correct in substance and in practice. It may be that the "algorithms" aren't the same as the ones in Photoshop's Composite Curve, but the effect is similar enough to warrant finding another way to work.

Cheers,

_andrew webb
___________________________________________________________________________
 
Re: Final Response to Your Summing-Up on the Camera Raw Dialogue
Posted by: Mark Segal
Tue Aug 21, 2007 4:46 am (PST)

Hi Dan,

I have read both of your closing messages on the Camera Raw debate, and I much appreciate the extensive amount of time you have devoted to this question, amidst all the other things that claim your time. In that vein, as we are summing-up on ACTL, I shall try to be concise, so this will not be an extended tit-for-tat response. I'll confine it to the key issues arising from these two posts, which are about commercial positioning, my stance in what I write, CR3.7, CR4.0 and the math, the process of publishing the papers, the availability of files, selection of images, comparing methods and results, and professional standards (including the Master Curve challenge).

Commercial positioning:
I hope you don't think that I have any commercial interest in anything I write, because I don't. I'm a pure consumer of Adobe's software, like tens of millions of other people, and I pay full price for it. I was very openly and publicly critical of the new print module in CS3, just as much as I am very openly and publicly enthusiastic about the new Camera Raw module. All this is just based on much experience using the stuff. No one pays me to write papers or take positions about anything. (That's not good for the relationship between time and cash flow, but it does underscore the fact that I am totally independent of commercial influence.)

Stance:
This leads to my stance in what I wrote. The only thing that motivated me to question your views of Camera Raw's capabilities and limitations is that those views and my experience collided. If you were not someone of your technical expertise in Photoshop, I wouldn't have considered such a reality check worthwhile, but when YOU say these things my reaction is "wait a minute, let's see what I'm missing here". So I took it upon myself to explore the issues and seeing the results, whatever you may think of the merits of the work, I thought it worthwhile sharing, and so did the publisher. In those papers, you will not find one reference to anything you ever did in Camera Raw 3.7 whether in PP5E or elsewhere. I responded to the statements you made about CR4 on List, and to avoid all possible ambiguity about what I was dealing with, I included a whole Annex of those statements. I told you off-list that I had done some work on three of the Chapter 16 images, and we agreed that before making any of that public, permission to use the images would be needed from two of the photographers and that if anything of this were published, in this case I would of course make it perfectly clear that some of the moves I can implement on those images in CR4 were not available to you in CR 3.7. So yes, I was taking issue with your views of CR4 as stated on List, but not with what you did in or about CR3.7.

CR3.7, CR 4.0 and the math:
That leads to the CR3.7 versus CR4.0 nexus. It was you who contended on List that the range-opening capability of CR4 remains as "crude" or "naive" or "unsuitable for professional quality results" as it was in previous versions because it relies on a master curve approach, and the master curve in CR and Photoshop are similar. This is the territory I was responding to in both papers. When we talk of adjusting contrast and the impact of contrast on colour in CR4, it can happen in different ways in four menus, and it is the combined impact on the pre-rendered image that matters. It is not only the curve; as well, even that curve is not programmed the same way in CR and in Photoshop. That is factual. But you did explicitly tell the List that you had worked out the exact math and programmed a spreadsheet. So you tweaked the curiosity of a number of people with this, therefore it was requested; the only reason I requested it rather more frequently than you seem to have liked is because you never responded to any of those requests. Had you said, for example, it is work in progress and you don't feel it appropriate to make it available just yet, I certainly would have understood that and not repeated the request, and I'm sure it would have been the same for the others. But you didn't. If you leave stuff like this hanging in the air, with the best will in the world, you have to expect it won't just go away.

Process of Publishing the Papers:
Regarding the sequence of sending you the papers and publishing them, yes - obviously we know you are busy and it takes time to read stuff and respond to it. There was an option to simply publish the papers with no notice. Michael Reichmann and I both thought, however, that we should extend you the courtesy of a preview and an opportunity to respond; therefore through me, he extended you an invitation to respond on Luminous-Landscape either along with publication of my first article or, recognizing you may not be able to do so just then, at any time you could manage it thereafter. You did not acknowledge that invitation. But you did manage to critique both papers on the List in response to posts from members, shortly after they were published on Luminous-Landscape, and even though you had not seen the files themselves.

Selection of Images:
One of the key things you critique is the selection of images: most recently the citrus image. I don't buy the half-pregnancy argument or that a black-and-white image would be construed to satisfy any of those criteria. That image was selected for the sole purpose of responding to ONE of your three categories, I said as much, and the characteristics of that image are what you described for that category. I'm not the only one to observe that. The fact that it also had a cast with significant detail was an added element to the challenge for that one category, not an attempt to kill two birds with one stone. If that wasn't clear enough, it is now.

Availability of Raw files:
As for the availability of the raw files, when I drafted the paper I knew of course that those files could be requested. It's a reasonable request. In fact I was expecting you to request them, but you didn't. OK, you're busy, there's only so much one should expect and that's your privilege. I don't yet have a website with server capacity to up-load large files and provide hyperlinks to the public for downloading. But as I've said before, the files are available on request, subject to minimal reasonable conditions that are normally well understood between people sharing files. Anyone who is interested enough in exploring what I did with those files and what they can do better with them is still invited to ask for the raw files and I shall provide them. I have no fear of being shown techniques that do better than what I did, starting from the raw files.

Likewise concerning the file for the B&W conversion, to set the record straight, I did not demur from providing the file. Off-List I sent you a JPEG of what the image looks like, saying on List I would provide the raw file to you once we agree on procedure, one key aspect of which is that you and/or your students use the same file I used so that we could compare like with like as between techniques of doing a B&W conversion. I still think that makes eminent good sense. You have in one email said you would prefer under these conditions to let the students bring what they will, while in one of these two posts you characterize me as exercising prudent judgment not to provide the file. I'm perplexed, but again, to keep the record straight I'm willing to engage in a sensible exercise comparing different techniques on a comparable basis, and I'd be very interested to see the comparative results.

Comparing Techniques/Results::
Look, there really is nothing personal in any of this, and I have no vested interests as I've said above. We're talking about what can be done with one piece of software versus another. The Camera Raw algorithms do not work the same way as Photoshop algorithms and the state of the data they are working on is different. As I think I've said, I've processed well over 1000 images in CR 4.0/4.1 of all kinds. I routinely combine various controls from the four menus. If anything, I would say my overall image quality is better (and processing faster) than it used to be. I simply have not come across issues of generating noise and artifacting from implementing the combinations of contrast and saturation or de-saturation adjustments that I describe in my papers - and I do know what my Canon 1Ds performs like and how to look at digital images as they progress through an adjustment workflow. Now, even if such accidents were to happen, it does need to be emphasized yet again: all of these changes are sets of meta-data instructions performed on the raw image data before rendering and the original raw file never gets damaged by these instructions, each state of the image is opened in Photoshop undamaged (but that the viewer may not like some of the renderings' appearance is a different matter altogether) and one may always revert to the raw image and change the instructions non-destructively. In this sense the raw converter is the true non-destructive image editor to the extent it can perform these editing functions.

The primary focus of my papers was to explore your statements about the limitations of Camera Raw. The issue of composite versus individual channel curves comes-up because of your critique of the composite curve. Where I considered it relevant to the basic issue of Camera Raw's functionality, I compared what CR does with its unique curve and what Photoshop does with the composite curve. I never intended or pretended that I would get into a whole further set of comparisons between using individual R, G, B channel curves versus either the Photoshop composite curve or the CR unique curve. I agree there are situations where using individual channel curves for making corrections is necessary and said so. We can get into an exploration about composite versus channel curves for particular kinds of images, but that is yet another talk-show and I don't think either of us have the appetite or the time for it just now.

Professional Standards (and Your Master Curve Challenge in CR vs. Photoshop):
To conclude, I must respond to the question of the standards to which these comparisons are made. You're right about the fact that different people have different standards. More than that, by virtue of training and experience, different people are attuned to seeing different things. You've complained over again that the PDF presentation isn't good enough for making statements about professional quality. I think the PDF illustrates the differences between image states and methods to which I was drawing attention. But fine, at some time when it is mutually convenient, let's take a number of raw images of my choosing and a number of yours (in both cases excluding images with multiple colour casts), arrange a group exercise with some professional retouchers whose skills you respect, and we'll have a derby (either virtual using a server or in situ) working them in Camera Raw versus working them in Photoshop, then we'll have a discussion about the quality of the comparative results. This could be a real interesting event. Short of doing something like that hands-on with the same raw files, it isn't appropriate to simply dismiss the other party's methods and results as not of a professional standard. Just to go back to the petunia, as I told you - printing Figure 22 would be a failure; printing Figure 23 would work. I prepare images that can print properly. I needn't tell you that printers have gamut constraints - and bright reds are notorious. I stopped the adjustments where I had an image that resembled the scene, conserved the detail and would print without nasty surprises. That is all part of standards. Maybe if it had to be pushed further in CR3.7 because the client wants more pop and all the detail I would be s.o.l. - I don't know because I haven't worked it in CR3.7. I keep coming back to the fact that I'm dealing with CR4.0 as an integrated package of controls, and I'm trying to demonstrate how it can do things which you are sceptical that it can do. So before we agree that I can't meet such and such a standard with it on that image, perhaps I should jump back into it and see if I can meet the hypothetical client's needs. Interesting idea. I'll try it.

Now, finally, (whew - longer than I expected) that raises your PDF. I've downloaded it and examined what you are saying about the three sets of images. I agree with you that the two bottom images on each page look VERY similar, but when I look VERY closely, I would say in all three the right bottom is ever so slightly more contrasty and more saturated than the left bottom, but we're talking quibbles. For the one on the third page, I have a slight preference for the CR rendition, because the facial shadow detail is a bit lighter and more transparent, but it's a quibble and could simply be a minor variance in curve drawing. That said, I think both bottom images on that page are over-contrasted.

This is where photographic taste intersects with the technical minutiae of how the images were produced. If I were working that third image, I would settle for contrast and saturation somewhere between the upper left and the bottom left. In the first image, highlights are badly blown comparing the upper left with either rendition on the bottom. Not having the original file I can't say for sure, but based on what I see in the PDF, I THINK (with some confidence) that if I had the raw file, I could render it using CR 4.1 more successfully than either of those demos. In the middle image I see no particular challenge one way or another. Almost anything could process that successfully.

But all that said about those three pages, I think there is one lingering misconception in the exercise as a whole - and I have alluded to it above - it simply isn't appropriate to treat Camera Raw as a Curve - it really was designed and should be used as an integrated package starting from the first menu, then moving from up to down and from left to right. This layout was not by accident - it sets down a workflow. It generally makes sense (though like for everything there are exceptions - I've processed some images in it that fare better by jumping straight to the Point Curve). So every image I process in Camera Raw is a strategic exercise: what do I want this image to look like; then, how do each of the tools in the toolset build to achieve what I want. Very often, by the time I render the image, it needs no more than a few things that are better to do in Photoshop: selective noise reduction when needed, sharpening using the Photokit Sharpener sharpening workflow and final tweaks in curves under softproof. It is to be hoped that CR will eventually allow one to load a printer profile and make all the adjustments in softproof mode. Just last night though, I was working on a slew of images of mosaics I shot inside the museum of Roman antiquities in Tripoli, Libya. The lighting there was not good for photography. Multiple casts and uneven. I needed extensive targeted adjustment layers, layer masking and gradients to rebalance them. This ain't for Camera Raw. Only Photoshop. I say all this to emphasize a very fundamental position that one should just use each of these programs for what they do best - I don't see much point trying to emulate the behaviour of a tool in the one application in the other.

On my side, I am also happy to bring this particular discussion to a conclusion as I also need to revert to other things, while leaving the door open for further comparative work down the road between Camera Raw and Photoshop; it is an interesting and important topic on which more can be done and said. So for now, while you wish us "Happy Curving", I shall wish you "Happy Rendering".

All the Best,

Mark
___________________________________________________________________________

Re: Final Response to Your Summing-Up on the Camera Raw Dialogue
Posted by: dave_cardinal
Tue Aug 21, 2007 2:26 pm (PST)

At the risk of adding another variable to the equation, given the differing color capabilities/response (gamut) of specific models of digital cameras, and the model specific coding/matrices in ACR (and other raw processors), people using images from a variety of cameras will see a variety of results about how effective any particular raw processor is.

To whatever extent this topic continues to progress, I'll be curious to know if some of the issues stated with ACR have more effect on one model/brand camera than another, and if that might account for some of the wide range of findings by the different but very experienced users on this list? Even something as simple as whether the file is destructively compressed or not (ala Nikon compression) could be a partial culprit if heavy processing causes artifacting.

It would also help individual photographers learn if a particular concern was relevant to them.--David Cardinal
___________________________________________________________________________

Re: Final Response to Your Summing-Up on the Camera Raw Dialogue
Posted by: Richard Wagner
Tue Aug 21, 2007 8:17 pm (PST)

David,

I think this is an excellent point.

Besides the "visually lossless" Nikon NEF compression that you referred to, there are undoubtedly countless other differences between RAW files from various models/manufacturers that may or may not come into play when comparing RAW processors or RAW processor functionality. I was recently told by an experienced Canon pro that setting white balance "in camera" gave him better results when using third-party RAW processors than setting WB after-the-fact in the RAW processor. This would be unexpected if the WB info was simply stored as metadata in the RAW file, as you should be able to get identical results with or without setting in-camera WB if that assumption were true. The implication, of course, is that setting WB in the particular camera model that he is using results in different in- camera pre-processing before the RAW file is saved. Undoubtedly manufacturers are aware of many of these "features" (and other useful data stored in makernotes) that are left to be discovered by 3rd party developers or photographers.

--Rich Wagner
___________________________________________________________________________

Re: Summing-Up on the Camera Raw Dialogue Addendum
Posted by: Mark Segal
Tue Aug 21, 2007 10:32 am (PST)

I realised early this morning, but too late to stop press, that I had omitted an important section of my "summing-Up" response. So here it is. This fits as the last paragraph in the section "Comparing Techniques/Results":

Dan correctly mentions that in the first paragraph of my first paper I said I would be comparing the use of the CR curve with that of individual channel curves in Photoshop, but then I failed to do so, insofar as where I did make Photoshop comparisons I used the master curve. Dan is partly correct and partly incorrect about this. I did say I would compare white balance (WB) and luminosity adjustments between them. In the final analysis, I did not compare WB between them. I just stuck with Camera Raw's WB and failed to make the point in the paper that it works so well (visually and by-the-numbers), there is no need to compare it with anything else. I did however say in the paper that such work can be done by adjusting individual channel curves as illustrated by Dan, Michael Kieran, Fraser/Blatner and others, and I provided references*. Over pages 14 to 18 of the first paper, I did compare CR adjustments with PS adjustments, and I agree I did not use individual channel curves. Again I failed to make the point in the paper that in my technical judgment - for what it's worth - I was not going to do better with three times the effort using R,G,and B curves. I agree this can be debated as I didn't nail it. However, in Annex 2, for that demonstration involving Lab space, I did indeed use all three Lab channel curves to prepare Figure 7, for comparison with Figures 3 or 5 done in CR.

*(As an aside, I have done some follow-up work on this matter, comparing the CR WB with the Photoshop Curves' grey balance eyedropper. CR WB does a better job - the algorithm is definitely different with noticeable impact on results.This would be useful material for the derby I proposed near the end of my posted response. I'm actually quite interested in further pursuing the issue of composite versus channel-by-channel curves several months down the road - it has not been exhausted either in my papers or in this dialogue. I expect what will come out of it is that doctrinaire positions one way or another don't have a place when correcting images in CR and PS.)

Mark Segal
___________________________________________________________________________

Range setting confusion
Posted by: Howard Smith
Wed Aug 22, 2007 7:41 pm (PST)

What, exactly, is gained by using ACR other than being able to open files that will not open otherwise (except in Lightroom, or other such specialized programs)? My own result, which can be uploaded to Files with David's permission, appears to me to be almost identical to the ACR result. Of course I'm not a professional photographer so my judgment may well be less than useful, but if all that is wanted is to obtain an image with good color and detail, why use ACR? There must be something more here than just color and detail. What am I overlooking? Or is that a question to which the answer would fill a book? If all that is really needed is a larger file containing more useful data, why not just open the image in ACR or Lightroom, export it, and edit it with Photoshop tools? It seems to me that white balance and everything else can be handled there. So far I've not experienced any problems with raw images opened in Lightroom and exported with no Lightroom edits.

Before anyone replies, let me assure all that this is not an attempt to generate controversy. My question is sincere, intended only to help clarify something that has long puzzled me.

Howard Smith
___________________________________________________________________________

Re: Range setting confusion
Posted by: Stephen Marsh
Thu Aug 23, 2007 2:40 am (PST)

Howard Smith wrote:

What, exactly, is gained by using ACR other than being able to open
files that will not open otherwise (except in Lightroom, or other such
specialized programs)?

Howard, these applications are converters/renderers – in the case of ACR, it converts grayscale raw camera sensor data into a RGB gamma corrected image suitable for display, output or editing. They render or `bake' photos from raw ingredients (linear grayscale data representing captured light levels) with various recipes (demosaic, white balance, exposure etc).

ACR/ALR now add further capabilities to work with non raw data and to save such into DNG format (previously solely associated with camera raw files only). This is what you were talking of with it being possible to edit say a JPEG file in ACR in CS3. This is very different from rendering a CR file. Notice how I said edit a JPEG and render a camera raw file, it is an important distinction. There may indeed be little advantage to the data of a gamma encoded file to be edited in ACR as opposed to Photoshop, that being said the workflow or controls may outweigh that factor.

So far I've not experienced any problems with raw images opened in
Lightroom and exported with no Lightroom edits.

No surprise here Howard, it can sometimes take a lot to break an image.

It depends on your viewpoint. If you are into "pristine data" then it is wise to use appropriate moves in the raw converter/renderer where possible, if they will not make your life harder or hurt the image for/in upstream edits or output. From a best practice view, this makes sense.

If one wishes to treat the raw rendering software like a scanner and create a flat image for doing all the work in Photoshop, then as you say many images may not visually suffer for the intended purpose or others and results may be very visually similar to the ALR process or better. This is not to say that the data is the same or may respond the same to future edits or output to other devices.

Bruce Fraser described an "inputcentric" and "outputcentric" viewpoint to scanning, which is what you are talking of here when using a camera raw file in ACR with flat settings and doing all the edits in Photoshop (in ACR one is editing in an indirect wide gamut high bit linear space, this is both part of the rendering process and an editing process, before the file is output/rendered/saved to a gamma corrected RGB file/colour space/bit depth).

All of this has been covered in recent links to PDF files and diagram uploads to the site.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: Range setting confusion
Posted by: "Denton Taylor"
Thu Aug 23, 2007 6:52 am (PST)

Hi Howard:

Speed, and the ability to apply a change made to one photo to multiple photos.

At 01:04 PM 8/22/2007, you wrote:
What, exactly, is gained by using ACR other than being able to
open files that will not open otherwise (except in Lightroom, or other such
specialized programs)?

Regards,

Denton Taylor
photogalleries at
www.pbase.com/dentontay/
www.dentontaylor.com
___________________________________________________________________________

Re: Range setting confusion
Posted by: Mark Segal
Thu Aug 23, 2007 6:53 am (PST)

Howard,

Yes, there are books answering all these questions - to be sure, and a lot of resources free of charge on the internet covering the same material. Don't worry about controversy, because it will always exist. Retouching and photography are separate professions so it is understandable that people whose lives are focused on retouching may not necessarily pick-up on the full photographic implications of these tools. I'll quickly lay-out a few of the most fundamental facts - as I understand them.

Firstly, photographers doing professional-level digital imaging - if they really know their craft - would ALWAYS prefer to work from a raw data file as long as time permits, because that is the source data which has the most information, the least amount of in-camera manipulation and therefore offers the most technical and artistic control in the least damaging way. The alternative is to let the camera bake a JPEG image. You have very little control over how the camera does that, depending on the range of user-selectable options, you largely accept the camera manufacturer's judgment about noise reduction, sharpening, contrast, hue and saturation, without exception you allow a tremendous amount of information to be discarded and the remainder then compressed, and you are confined to an 8-bit depth. When you receive this image onto your display your options for fixing what you don't like are severely constrained because the image will literally fall apart much more readily than would happen with a raw file. The JPEG format was developed for speed and convenience (which many people need, so it's fine in those circumstances), but it is not for exacting quality and preservation of editing head-room.

Secondly, once you make a raw format exposure it must be de-mosaiced and rendered into a multi-channel format with a colour space and bit depth using a raw converter. That's the most basic function of a raw converter. But the development of raw converters over time has seen more and more image adjustment capability built-in prior to rendering the image - this because the more adjusting one can do non-destructively the better. The technical approach for doing this is the creation of meta-data instructions "on top of" the raw data which operate on the image at the time it is rendered, but does not alter the original raw image data itself. Therefore what you captured is always there in its unaltered state if you go back into the raw converter and reset the instructions to the start point. This is the only truly non-destructive stage of image adjustment. Many people would consider it just as important to have at least one record of their original capture to the same extent they considered it important to safeguard their negatives and chromes during the film era.

Thirdly, at the present state of programming - at least within the Adobe family of software - there are things you can do in the converter that cannot be done at all or as easily in Photoshop. Some examples: (i) One of the key things is the recovery of clipped highlights, as long as one or two channels of information are not totally clipped. (ii) As well, the white balance eye dropper in Camera Raw 4.1 (at least from my observation) does a more accurate overall job of correcting single colour casts than using the grey eyedropper in the Photoshop Curves dialog, which is the closest "equivalent" tool between the two applications. Doing the white balance in Camera Raw is a real work-saver. (iii) The Camera Raw Curve preserves hue in a way that the Photoshop composite curve in principle does not.

Fourthly, and here is where we move from fact to judgment, the adjustment capabilities built-in to the new Camera Raw 4.1 - arguably - for a great many imaging situations (but not all) are capable of very high quality renditions of contrast, of image detail, and of desired hue and saturation more easily than achievable in Photoshop, provided one starts with a raw file. As well, all these instructions operate as a set of rendering instructions, so the image opens with all this user-controlled adjustment built-in and without having altered a rendered pixel. You can also create this adjusted version of the raw data as a "Smart object" exported to Photoshop, so once there, if you wish to go back to perform editing functions using the raw converter, you can do so from Photoshop by clicking-open the Smart Object layer, and those changes will "convey" directly into the Photoshop document - neat, eh?

Now, moving from facts to observation, I think the new parametric curve in Lightroom and Camera Raw produces different effects on luminosity than does the Photoshop composite RGB curve. I know Dan Margulis disagrees with this, but that discussion needs to be underpinned with more empirical research that is properly structured.

In a nutshell, the raw converter is all about preservation of image data, artistic and technical control and image quality. Finally, apart from some user interface refinements found in the Lightroom Develop module, that module and Camera Raw 4.1 are pretty much identical.

Mark Segal

PS. Just for clarity I should note that "Red Flower Fig.21 in its raw state is my image - of which David Marley extracted a compressed, low-res Acrobat formatted version from my paper.
___________________________________________________________________________

Re: Range setting confusion
Posted by: dave_cardinal
Thu Aug 23, 2007 8:04 pm (PST)

Just to lay some general background in here for folks who aren't familiar with what White Balance started out doing (clearly tools labeled WB like the current one in ACR have grown more and more complex), White Balance on early cameras & raw converters was very simple. It "just" multiplied the Red and Blue primaries by a value to adjust the image as if the light source had been some other Kelvin temperature (on the simplistic assumption that light sources ranged from warm to cool but were fairly homogeneous otherwise). Nikon even conveniently encoded the two multipliers in the images. We published a table of them in _The D1 Generation_ in 2001.

Quickly though camera makers (and software developers) started to add complexity to WB corrections to adjust for (for example) fluorescent fixtures not having all the frequencies. So on more recent Nikon Camera (probably Canon also, I just haven't done the experiments) Fluorescent and even Incandescent settings weren't just simple math, they actually had some knowledge of the light source specified (Nikon offers a couple Fluor. Setting for different filaments). Similarly the flash setting started to get special treatment because the Nikon speedlights are slightly different from the equivalent Kelvin temp.

Now the camera vendors even mix in a Hue adjustment (and Adobe has gone to two sliders) so you can officially tweak the WB in two dimensions, not just the warm/cool of red/blue that dates back to the color temp on film.

So why is this important? Well to me it is because I really want my WB tool to be made to do WB. It should (and I find that Nikon Capture NX and ACR 4 do, at least to my liking) be able to correct for light as well or better than me fiddling with curves. So I _always_ do my WB correction on the Raw image with a WB tool meant for it (or in the camera if I'm shooting JPEG).

That doesn't mean it is the right tool to correct for some other type of cast in the image, or even for multiple casts, but if what you want to do is adjust for a light source of a different temp then that's certainly what WB should be able to do well.

--David Cardinal
___________________________________________________________________________

Re: Range setting confusion
Posted by: Mark Segal
Fri Aug 24, 2007 11:16 am (PST)

Stephen,

Maybe this is just a quibble about words, but sometimes the nuances of vocabulary can lead people into thinking inaccurate things, so I think is useful to characterize these processes a bit differently. Normally, we don't speak of "rendering" and "baking" in the same terms. This isn't being facetious, but when you bake a cake, you cannot thereafter unscramble the ingredients. When you "render" a raw image you can go back and unscramble the ingredients very easily - simply by changing the meta-data instruction set. The raw data itself never changes. What's more, you can embed the raw conversion settings in a Smart Object and make these changes in Camera Raw right from Photoshop. Then we should distinguish between "recipes" and "ingredients" - demosaicing is indeed a recipe - cast in silicon, we have no control over it. White Balance and Exposure on the other hand are "ingredients" - season to taste, and recompose as often as you like.

Now - can you wreck the cake with all this? NO. You might not like the result you get from the ingredients you used, but you CANNOT damage a raw image. It's the negative. You can always go back to the same unchanged information and alter the meta-data instructions (ingredients) - according to the late Bruce Fraser - with much less risk of data damage than doing the same in Photoshop. So your bottom- line advice to Howard on making the best of pristine data is exactly how I view it. It's often not so much about whether visible damage occurs with a current set of edits, but minimizing the risk of such damage, and maximizing the scope to repurpose the file for possibly more demanding output conditions sometime in the future.

But beyond that important factor, there are new algorithms in Camera Raw 4.1/Lightroom 1.1 which don't exist in Photoshop and are seriously good tools for achieving certain objectives more easily, or opening possibilities that were previously considerably harder to do.

Mark Segal
___________________________________________________________________________

Re: Range setting confusion
Posted by: Stephen Marsh
Fri Aug 24, 2007 6:07 pm (PST)

Mark Segal writes:

Stephen,
Maybe this is just a quibble about words, but sometimes the nuances
of vocabulary can lead people into thinking inaccurate things, so I
think is useful to characterize these processes a bit differently.

Fair enough Mark, whatever works for you, I am not really into these metaphors and bake is probably the most common one used to describe the in-camera process of CR>JPG/TIF, the raw data is untouched and the conversion could be undone – if the raw data was not automatically deleted in-camera.

I have some similar nitpicking over a recent post that you made and would like to clarify what you are saying, but that will have to wait until I have more time. After all, where would this list be without debate over such nitpicking. With the help of the older ACR chapter download posted by Rich and the diagrams uploaded by Rich on CR pipelines, I feel that I can make more sense with my questions and need for clarification, which your post reinforced.

Normally, we don't speak of "rendering" and "baking" in the same
terms. This isn't being facetious, but when you bake a cake, you
cannot thereafter unscramble the ingredients. When you "render" a raw
image you can go back and unscramble the ingredients very easily -
simply by changing the meta-data instruction set.

Agreed, but the rendered/processed file output from the converter is converted/created/rendered/output/baked, whatever term may be used. If you prefer render than bake, then fine. The camera renders (oops, I meant bakes) a file too, one just has a lot less control over the render.

If what you are saying is that in these discussions it is best to say "render" for the process of CR file to RGB output - and only to use the term bake to describe an in-camera process where the CR data is deleted, then I can sort of see your point about 'bake'. I use that term perhaps more loosely, to describe any CR>RGB output process, be it in-camera or on the computer using ACR/ALR etc.

The raw data itself never changes.

I dont' think that I ever said it did.

We generally agree on the major points, as is often the case with various discussions on this list – but the sticking points are often minor and tend to drag out into long debates. I am sure that will not be the case here.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: Range setting confusion
Posted by: Mark Segal
Fri Aug 24, 2007 6:37 pm (PST)

Stephen,

Indeed, yes, this is exactly the point. Good generic distinction of two very different situations.

Mark Segal
___________________________________________________________________________

Re: Range setting confusion
Posted by: Richard Wagner
Fri Aug 24, 2007 6:47 pm (PST)

On Aug 24, 2007, at 6:09 AM, markds0 wrote:

Then we should distinguish between "recipes"
and "ingredients" - demosaicing is indeed a recipe - cast in silicon,
we have no control over it.

Mark,

I agree entirely with your post - except that statement. Demosaicing is done in software, and there are a lot of algorithms to do it, so in that sense it's not "cast in silicone." DCRAW actually lets you choose which of 4 algorithms to use:

- q 0
Use high-speed, low-quality bilinear interpolation.
-q 1
Use Variable Number of Gradients (VNG) interpolation.
-q 2
Use Patterned Pixel Grouping (PPG) interpolation.
-q 3
Use Adaptive Homogeneity-Directed (AHD) interpolation.

There is a great overview here:
http: //scien.stanford.edu/class/psych221/projects/99/tingchen/

--Rich Wagner
___________________________________________________________________________

Re: Range setting confusion
Posted by: "Ron Kelly"
Sat Aug 25, 2007 7:41 pm (PST)

David:

That's basically fine, but not to split hairs I have a slightly different perspective.

The white balance adjustment as you describe will be done visually, no? Therefore, it only "gets you close" to the final result. You're still probably going to color correct in Pshop with the info pallet for highlight color anyway.

Why is that so different than setting the WB manually (in camera), or allowing the Auto function on the camera to do it's thing? You'll be close, and you'll still color correct in Photoshop.

Adjusting white balance with the monitor as your guide is fine, and maybe many are much better than I am at it. However, do you trust you profile adjustments to the visual guide in "Display Calibrator Assistant" (Mac) or the equivalent in Windows? I bet not.

So tweaking the white balance 9 ways from Sunday in the Camera Raw tool is nice, but it's the rough cut anyways. Final shaping is still to come with precision tools.

Ron Kelly
___________________________________________________________________________

Re: Range setting confusion
Posted by: Stephen Marsh
Sun Aug 26, 2007 7:12 pm (PST)

Mark Segal wrote:

This is very simple and has only a little to to do with curves. Make
a photograph that has well-known colours and make sure the MacBeth
ColorChecker is in it, such as the citrus example in Annex 3 of my
paper.

Not having immediate access to the image above (yes, I know that I can ask), I would have used the one found below (19mb).

http: //s179771984.onlinehome.us/RAWpository/images/nikon/D2X/1.01/_DSC0030.NEF

On a side note, if you would like to do your bit for the imaging community, submit a rights free image of a colour target and some fruit or a boring outdoor shot or something meeting the guidelines, as it would appear that this project is not being well supported and the stated purpose seems noble enough (I am getting around to doing this, so I hope not to be a hypocrite):

http://www.glasslantern.com/RAWpository/

Regards,

Stephen Marsh.
___________________________________________________________________________

Re: Range setting confusion
Posted by: Mark Segal
Mon Aug 27, 2007 1:00 am (PST)

Stephen,

I'd like to be helpful, but as I mentioned in one of my responses to Dan, I am not equipped with accessible server space just now for up-loading files. If I remember correctly was it David Cardinal who said he could offer space? - so that may be a solution.

Also, before getting into that I don't understand some of your message, i.e. what "project" is not being well-supported? What guidelines are you referring to?

Mark Segal
___________________________________________________________________________

Re: Range setting confusion
Posted by: Stephen Marsh
Mon Aug 27, 2007 1:24 am (PST)

Hi Mark, I understand about the server space, that was not a dig, just a helpful note to anybody wishing to take you up on your offer but not wishing to wait.

Also, before getting into that I don't understand some of your
message, i.e. what "project" is not being well-supported? What
guidelines are you referring to?

The site and guidelines were linked directly after that text, which was indicated via a colon at the end of the paragraph inviting submissions (I have no affiliation with this website). The website in question is the Raw Repository (where one can downloaded a CR file of a test chart and fruit to play with colour balance or anything else):

http://www.glasslantern.com/RAWpository/

I said it appears that they are not well supported, judging by the limited amount of camera model files available - compared to all the cameras on the market capable of shooting a CR file.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: Range setting confusion
Posted by: David Cardinal
Tue Aug 28, 2007 1:35 am (PST)

Ron--We probably need to differentiate between different customers & different workflows here. Many photographers aren't really interested in mathematically accurate reproduction, but instead pleasing reproduction. So visual is what it is all about for them (which includes me for most projects).

I do spend a lot of time and money to get my prints to match my monitors as best I can, because I use those monitors to judge what I consider visually pleasing before printing, and if I get it right I save myself time and money.

But I almost never rely on pure numbers to set white balance with wildlife photography. I've written a couple articles explaining why but overall it is because the emotion of the scene is affected by the WB and that's an artistic decision. To me that is one of the coolest things about Raw is that I can do that setting fairly flexibly "after the fact".

That doesn't mean I don't tweak colors after that (I love Curvemeister!!) but I don't do it to set the WB. That's already done by the time I get it into Photoshop. But certainly I can imagine lots of different approaches, especially if you need to precisely color match your subjects in your prints.

--David Cardinal
Cardinal Photo / Pro Shooters LLC
http://www.cardinalphoto.com
http://www.nikondigital.org
http://www.proshooters.com
___________________________________________________________________________

Re: Range setting confusion
Posted by: Richard Wagner
Tue Aug 28, 2007 7:34 am (PST)

On Aug 27, 2007, at 4:58 PM, David Cardinal wrote:

But I almost never rely on pure numbers to set white balance with
wildlife photography. I've written a couple articles explaining
why but overall it is because the emotion of the scene is affected
by the WB and that's an artistic decision. To me that is one of
the coolest things about Raw is that I can do that setting fairly
flexibly "after the fact".

I agree 100%. WB can be used like a set of filters, and adjusting WB can be used to change the mood. Cool, cold, warm, whatever you'd like, including multiple interpretations (renderings) from the same capture, and all without damaging the image. (Picture the guy in waders fishing - blue light, or orange? Both can be dramatic, with opposite moods.) As Andrew Rodney used to say here over and over, we seldom want scene-referred colorimetrically correct images - we usually want pleasing images. One of the easiest ways to control this is with white balance. It is also one of the reasons that it is often unwise to give your RAW files to someone else to convert, as their rendering might be unlike anything you imagined. If you do the conversions yourself, you can play with WB until you get the look you want. Most of us simply use the WB measurement as a starting point (unless accurate color is a requirement). This is a huge advantage to shooting RAW.

On the other hand, if you want accuracy (visual or otherwise), there is no substitute for accurate WB adjustment in the RAW converter and not in Photoshop, because WB adjustment is an exclusive RAW converter tool. It is also not possible to "ignore" WB when converting images, as whether it is ignored or not, it is set to something. If it is set wrong (wrong look or wrong numbers), it will be essentially impossible to undo later in Photoshop, because it does not produce a simple color cast that is easy to correct with Curves. To the contrary, everything gets whacked.

White balance is a fundamental operation in RAW conversion and digital capture in general, and you simply can't ignore it or try to do it later in Photoshop and produce successful images.

--Rich Wagner
___________________________________________________________________________

Moderator notice: Range-Setting Confusion Closed
Posted by: Dan Margulis
Sun Sep 9, 2007 7:59 pm (PST)

I have just returned from three weeks in the west. Before leaving, I requested that the Camera Raw-related threads be brought to an end. Stephen Marsh echoed this request about a week ago, to little effect. Accordingly, these interlocking threads, which are now, believe it or not, three months old, are hereby closed--no further posts on any topic relating to aggressive-vs.-conservative use of ACR will be accepted.

This list targets high-end image manipulation. Anybody claiming that ACR does it better must exhibit not just a knowledge of how ACR works, but what the realistic alternatives (not master curves, not eyedroppers to set neutral points) are in Photoshop. This knowledge has, IMHO, been absent in both respects so far in the posts of those who have advocated aggressive use of ACR.

With respect to the range-setting routines of ACR, I have shown that they can be emulated closely in Photoshop using a master curve. I have not indicated where that master curve is applied--I leave that to the advocates to figure out, and demonstrate to the list. With respect to the White Balance command, I would ask that they either show an emulation in Photoshop similar to the one that I did for the Tone Curve command, or else at least address some of the following questions about behavior:

*All attempts to regulate highlight balance tail off as the image gets darker. Where is the endpoint of the effect of the ACR White Balance command, and how does it tail off?

*What is the effect on the impact of the White Balance sliders if the user refrains from setting range with other commands, as opposed to attempting to set highlight and shadows to specified values?

*In PP5E, I demonstrated that the use of the ACR White Balance sliders can change image luminosity markedly as well as color. When does this occur, and why?

Yes, I know the answers to these questions, because I have tested the product. The least we can ask of those who advocate aggressive use of it is that they do the same. If there is a good faith effort to answer them, then I would say the answers would be of relevance to list members and discussion would be welcomed. If, however, we have nothing more that unsubstantiated claims of superiority for aggressive ACR correction, then I would say we have had enough of that kind of thing for at least another year.

I will shortly be posting comments about what took place at PSW.

Dan Margulis
___________________________________________________________________________

Re: Moderator notice: Range-Setting Confusion Closed
Posted by: Stephen Marsh
Mon Sep 10, 2007 4:09 am (PST)

Just to satisfy some list members, I would like to clarify the thread closure by Dan.

Dan beat me to the punch when he returned today, I was going to put a close on the thread today, in view of recent posts.

When I commented that it appeared that the thread had reached it's natural conclusion and that we could all move on, I was asked off-list not to close the thread so soon (regarding recent white balance posts). I left the thread open, to see if any useful last minute closing statements would come forth. Sadly the thread continued much like it had earlier.

For those complaining of intentional double standards in letting some messages through for approval when others have been halted or reworded due to their wording or content - this is not the case. What one may find is natural variation in the moderation as it done by various individuals in different time zones and different countries.

I personally think it is best if all posts are not approved if they contain *any* off topic or potentially flammable comment/remark etc. This may mean that 90% of the regular list traffic is rejected until people get the hint that this is not the Adobe forums or that past tolerated posting behaviour on ACTL will no longer be tolerated.

Sincerely,

Stephen Marsh.
List Moderator.
___________________________________________________________________________

Photoshop Curves vs. Lightroom Curves
Posted by: "Richard Wagner"
Mon Oct 8, 2007 9:28 am (PST)

(INSERTED BY MODERATOR)

After several unsuccessful attempts to close a thread that had meandered for three months, on 9 September we announced a firm close to the thread and stated that we would not accept further postings on the question of whether files opened in Adobe Camera Raw should be corrected aggressively or conservatively before entering Photoshop proper, *unless* those advocating an aggressive approach demonstrated knowledge of or a serious attempt to study how the ACR commands compared to alternatives in Photoshop proper.

The post below is an attempt to reopen the thread. As the author of the article being cited has a commercial relationship with Adobe, we have held off approving the post until we had a chance to review it to see whether it contains the type of technical information specified above. Happily, it does: the author tries to explain how to emulate the function of ACR's Tone Curve command in Photoshop.

I will probably have a brief response. If anyone else cares to comment on the article, the discussion is of the Tone Curve command ONLY.

Dan Margulis
(/MODERATOR INSERT)

Martin Evening has posted an interesting article on LR News about the
similarities and differences between Curves in Lightroom and Curves
in Photoshop. For those on the List interested in Curves, it's an
interesting and informative read.

http: //lightroom-news.com/2007/10/04/lightroom-versus-photoshop-curves/

--Rich Wagner
___________________________________________________________________________

As can be seen, between August 28 and the official closure of the thread on September 9, there wasn't a single post that I would consider worth archiving. Presumably because of fatigue with the topic, no list member responded to the invitation to discuss the Evening article cited above. Therefore, although I had prepared them, I withheld my own comments at the time. It should be noted, however, that the Evening paper confirms completely (with tests similar to those shown in my own presentation)that the Tone Curve commands in Lightroom and Camera Raw are nearly identical to the master curve commands that have been found in Photoshop since the beginning of time--commands that serious users, even in 1992, would have known to avoid.

In the interest of completeness, I append the comments  that I did not post at the time. --DM

A recent link to an on-line article by Martin Evening bears on the very long ACR thread that we recently closed. This piece is the type of informed commentary that was missing in much of our thread. It discusses how the Camera Raw Tone Curve compares to moves in Photoshop proper and shows how to create results that are very similar in either, as I had previously demonstrated was possible in
http: //tech.groups.yahoo.com/group/colortheory/message/18453

Because it verifies what I had been saying throughout the thread with respect to master curve behavior, a few comments are in order.

I. METHOD OF COMPARISON.
Martin's method is the correct one of attempting to match a known result from ACR by using similar methods in Photoshop. His results vary only slightly from my own. He demonstrates that when a master curve in a shape very similar to the Tone Curve is applied in Photoshop, the result is difficult to distinguish from the ACR result.

He notes that certain variations result from a conscious decision to change some hues. Although he presents a table showing some of these variations, he does not incorporate them as part of his Action, as I had done.  However, his example image verifies what I had said in this regard. It shows a parrot with a yellow chest and blue wings. He presents a version produced by a Tone Curve in ACR vs. a version produced by a similar master curve in Photoshop.

I described the hue changes in
http: //tech.groups.yahoo.com/group/colortheory/message/18448
"[These hue changes] are attempts to improve fidelity in certain color ranges. Unless you are doing something unusual, such as applying a strong inverted S curve, the likely noticeable areas would be that light cyan-blues, such as found in many skies, become darker and shift toward blues, and dark orangy-reds becomelighter and shift toward yellows."

Martin's sample image confirms this point. In the version done with Photoshop master curves, parts of the chest are slightly darker and more orange than in the ACR version. Otherwise they are very close indeed.

Martin's master curve is slightly modified from mine. I had indicated that my Photoshop master curve was identical to the ACR tone curve; his has some extra holding points plus a supplementary contrast curve. The difference is due to a slight error that does not affect his basic conclusions. When I showed my Action, out of disgust with the low technical level of the thread, I said that I would not specify in what RGB definition the Photoshop master curve needed to be applied and would leave that to the advocates of aggressive ACR correction who had not yet figured out that  the command was even a master curve at all.

The correct answer is that it needed to be applied to an RGB workspace using the same primaries as ProPhoto RGB, but a gamma of 2.2. (The gamma of "real" ProPhoto RGB is 1.8.) Martin applied his curve to "real" ProPhoto RGB, hence the need for extra points. The omission does not affect his conclusions. For practical purposes on this list, his explanation of how the ACR Tone Curve command functions may be taken as correct.

II. REFUTATION OF CLAIMS BY OTHERS.
This article torpedoes three of the bogus claims that pervaded our list over the course of our thread. First, both Martin and I have shown that we can replicate ACR's Tone Curve behavior in Photoshop. It was repeatedly claimed that this was not possible because of the extreme sophistication of the ACR method and that nobody could possibly "reverse engineer" it as Martin and I did.

Second, opening range in ACR, we now all agree, causes an increase in saturation in lighter colors, whether we like it or not. It has been repeatedly suggested that this is due to a conscious decision by the programmers. Both Martin's and my results indicate that the increase in saturation is accidental, the inevitable result of applying a master curve to an RGB file with ProPhoto primaries. Applying the master curve in Photoshop proper results in a nearly identical saturation increase. The only deliberate actions in ACR to alter the result of the master curve are the hue changes noted above.

Third, when I suggested on this list that, contrary to the conventional wisdom, the behavior of ACR in setting range is that of a master curve,  it provoked the following responses, among others:

1) "This is pure, unsubstantiated, meaningless horse manure...To state that *
any* RAW processor is equivalent to using "master curves" in Photoshop is sheer nonsense."

2) "You keep saying this kind of unsupported nonsense...This idea of a master curve is absolutely not how the product handles the raw data, not even close. You really need to have a conversation with Thomas Knoll, Mark Hamburg (if they'd even talk to you) or read Bruce Fraser's Real World Camera Raw. You are either ill informed about the processing pipeline or you're making this stuff up!"

3) "Facts have been presented that directly contradict your supposition, and you keep saying the same thing. You have used a
single image and a single metric to base your claim that it "uses a master-curve approach", that being the saturation change that occurs.
There are other behavioral differences between curves in ACR and Photoshop that are considerable, and make them significantly different, and
makes your characterization that CR tone curves works in the way you describe factually incorrect. The exact mathematics of this function are in the DNG SDK. It does not jive with your description, at all"

4) "Camera Raw doesn't use a Master Channel adjustment like Photoshop."

As both Martin's and my results indicate, not only does Camera Raw use a master channel adjustment *like* Photoshop, but the master channel adjustment is nearly identical. IMHO, a competent person could demonstrate to his own satisfaction that the ACR Tone Curve is a master curve in around 15 minutes. And yet, we went on for nearly three months arguing about whether this most critical function of ACR, was or it wasn't. It was for this reason that the moderators finally announced in disgust that we would not accept further advocacy for any aspect of ACR unless the advocates demonstrated that they had attempted to figure out what its commands actually do, as Martin Evening has.

III. HOW SOPHISTICATED IS IT?
Since there is no longer any dispute as to what the command does, it is well to see how it compares with past practice. My samples and Martin's verify that it is a master curve, not a single curve that acts on some type of averaged luminosity values and then establishes color. Further, we have shown that the differences between the ACR Tone Curve and the Photoshop Curves command are inconsequential--far closer than, for example, Curves and Levels.

Master curves apply identical procedures to each channel.  Professional users have known about the deficiencies of this approach for a very long time.

I did not use Photoshop 1 personally, although I believe that Curves was there. I did use Photoshop 2.0 in 1992 and so can verify that a similar, nearly identical, master curve was available.

However, sensible Photoshop users of 1992 would have known better than to use a master curve for any major maneuvering. Master curves create color issues. They exaggerate existing casts, and can cause unwanted saturation gains at the endpoints. More importantly, they damage detail in objects with a pronounced color. This makes them particularly unsuitable for photographs such as the one of the parrot presented by Martin. The bird has a very yellow chest, and brilliant blue plumage elsewhere.

In the red channel of this image, the yellow areas are extremely light. The blue areas are very dark. In the green channel, both areas occupy the midrange. Therefore, critical detail resides in every range of the curve, in one channel or another. Any curve, master or otherwise, always emphasizes detail in certain areas at the expense of others. Using channel curves, experienced users allow the damage to fall in unimportant areas of channels. But with a master curve that applies the same correction to all channels, this approach does not work. It this parrot picture is not possible to apply a master curve without damaging detail somewhere. It can only be retained with channel curves, or with a single curve the applies only to composite values. Although this raw image was not made available for download, the damage to the red channel is clear from opening the low-resolution version attached to Martin's article (see discussion below).

In ACR, the Exposure, Brightness, and Tone Curve settings are all master curves; there is no more sophisticated method available. In comparing the Tone Curve command to what an unsophisticated user would have done in 1992, pluses and minuses appear.

*As Martin indicated, the ACR master curve is applied using the primary color definitions of ProPhoto RGB, an ultra-wide gamut space. In 1992, a master curve would have been applied using narrow-gamut primary definitions, similar to those of Apple RGB. The 1992 method would have been better for color issues. Casts and faulty saturation are exaggerated much more in ultra-wide RGBs. OTOH, while there would be loss of detail either way, it might be easier to recover from the damage in ACR than in 1992. In ultra-wide RGBs, the weakest channels are darker and more detailed than in narrower-gamut RGBs, so they are not as susceptible to being blown out altogether.

*The hue adjustment that has been added to ACR is sometimes helpful and sometimes not. My tests indicated that its adjustments in blues were pleasing, but in yellowish-reds, not. In this parrot image, I prefer the version done by Photoshop master curve to the one done with the ACR tone curve, although neither is professionally acceptable. Areas of the chest in the Photoshop version show a transition from yellow to orange that is more pronounced than in the flatter-looking ACR version.

Taking all the above pluses and minuses into account, I would say that on the whole the mandatory ACR range-opening routines are slightly worse than what an unsophisticated Photoshop user might have done out of ignorance in 1992, however, it is arguable that they are as good.

IV. ANALYSIS OF IMAGE.
Martin makes the argument that is familiar to readers of our very long thread. He states that applying the Tone Curve in ACR is better than applying a master curve in Photoshop, and also better than applying a master curve and then fading back to Luminosity mode. In fairness, the point should also be made that it is better than applying Gaussian Blur to the image at a radius of 25.0 pixels, or better than activating a painting tool and painting a large letter X across the parrot. A professional retoucher would be as likely to use any of these four approaches as one of the others.

The red channel of the parrot image has not just been damaged--it has been obliterated in critical areas. Around a quarter of the parrot's yellow chest is missing entirely from the red channel. Significant parts of the wings are completely plugged, printing as solid black. Consequently, the image is not professionally acceptable. The colors are pretty but the detail is gone.

We saw a similar situation recently in an image of pumpkins posted by David Cardinal. Because there was no detail in the critical red channel, the pumpkins were just a mess or orange rather than the three-dimensional objects normally adorning front porches on Halloween. As shown in a corrected version by André Dumas, constructing a proper weak channel results in a massive quality increase.

If we were presented with Martin's JPEG image as a fait accompli, we would correct it by using a modified copy of the green channel as a base to blend detail back into the red. If instead we were given the raw file, and someone held a gun to our heads and demanded that we apply the damaging Tone Curve, then we might attempt to counter-correct by decreasing saturation, and then, granted that the parrot needs to be fully saturated, counter-counter-correct with one of the ACR commands introduced in Photoshop CS3.

The chances are that neither of these approaches would be able to fully compensate for the application of the master curve. However, even if they could, it would make little sense to use either one in preference to opening the image in a raw module with superior channel control, or just opening conservatively in ACR.

V. SUGGESTED FIXES.
As noted above, professional users have known for a long time that master curves should be avoided. In the first edition of Professional Photoshop (1994) an entire chapter is titled "The Unwanted Color and Why We Want It." It discusses how the weakest channel in a strongly color object is of critical importance in defining shape. It points out that we normally need to *intensify* detail in that channel by such means as channel blending, let alone damage it by means of a master curve.

Similarly, in 1996, my column "Defanging the Curves Vampire", which was aimed at beginners and is often used as a university text, has en entire section entitled "Beware the Master Curve!" explaining the detail loss that it guarantees.

In the last few years, the importance of adding localized contrast to images has been strongly emphasized. The concept of high dynamic range, and my own "picture postcard" workflow, depend heavily on maximizing contrast per channel. Photoshop's own curve structure is rather long in the tooth for this purpose. Many more sophisticated methods are available. AFAIC it is regrettable that the contrast-opening structure of ACR should not be made *better* than Photoshop proper's, let alone returning us to a way that was second-rate in 1992.

Martin offers the suggestion of adding a feature to the current ACR Tone Curve so that users could be able to control the increase in saturation. That is, he would allow users to choose a point anywhere between the full saturation gain of the ProPhoto master curve at one extreme, and no gain in saturation at all on the other. While this would be an improvement it does not address the main point of detail loss.

If such a patch were to be considered, the appropriate way would be to calculate the curve internally twice: once as it currently does, to be used for color values only, and a second time applied to a luminosity channel derived from the RGB channels (NOT a master curve applied to all three and then averaged to find luminosity). Then, the two curves could be used to form a unified output, and THEN Martin's suggestion of a slider to control saturation gain would be valuable.

This method works because an artificial channel based on luminosity has strongly-colored areas such as those of the parrot in its midrange--neither light nor dark, as color is not relevant in such a channel. Therefore, an S-shaped curve helps contrast in *all* colored areas of the image, instead of blowing out critical detail in the strongest and weakest channels, as a master curve does.

The luminosity-plus-saturation approach described above is still quite crude--but it would be infinitely better than the way ACR does things now. It would never be worse, sometimes the same, and frequently better. It would be unlikely to damage the image enough to hamper future correction, the way ACR's master-curve approach currently does. And, in some cases it would be quicker and easier than corresponding moves in Photoshop. But it would still be a ways from fully meeting the needs of professionals.

Several list members have recently suggested ways to improve curve functionality. Mike Russell's new release of Curvemeister (I haven't seen it myself) purportedly includes some sophisticated masking technology to modify what the curves do. In a raw module, developers are not limited to traditional RGB channels--they can adopt CMYK, LAB, or other models for the curves and then output in RGB without damage. Adding any of these more modern methods either to ACR or Photoshop proper (or preferably both) would certainly be welcome.

Meanwhile, Martin Evening should be commended for having stated what the Tone Curve command actually does in comparison to methods available in Photoshop, thus allowing people to draw their own conclusions about its merits.

Dan Margulis