Dan Margulis Applied Color Theory

CS3's B&W Command for Contrast

CS3's new B&W tool used for Contrast in color images
Posted by: "williamtheis"
Tue Aug 7, 2007 2:49 pm (PST)

In CS3, under Image>Adjustments>Black & White (shortcut: ALT-SHIFT-CNTRL-B) resides Adobe's new Black & White conversion filter. What is the collective thinking about using this for image contrast control? I add a layer and apply the filter, able to observe the B&W image until it is as desired, then change the layer to Luminosity mode.

It really helps to study the image since the Black & White filter has many knobs. I analyze similar to Dan's PPv5 chapter 7 "Keeping Color in the Black & White" to guess beforehand the "friends" and "enemies" that I wish to either enhance or kill. This to me seems a logical alternative to the channel blends since you can readily visualize the results.

Comments?

Bill Theis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Stephen Marsh
Tue Aug 7, 2007 5:44 pm (PST)

Hi Bill, or one can switch to luminosity mode first and evaluate in colour the luminosity blending effects as you move the sliders to add drama to sky or separate foliage from backgrounds or whatever the task may be.

I mentioned this technique on list when the CS3 beta was out and nobody replied:

http: //tech.groups.yahoo.com/group/colortheory/message/16406

I only played during the beta and not since. I found it interactive and fast but easy to create posterization and artifacts with some combinations of sliders and image colours/tones etc. I found doing this in ProPhoto (8bpc or 16bpc did not seem to matter) gave smoother results with less posterization.

http: //members.ozemail.com.au/~binaryfx/howto_lumabw.html

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: "williamtheis"
Tue Aug 7, 2007 11:24 pm (PST)

I think the technique has real promise and am surprised at the lack of response! Glad you tipped me off to the other links, which I will study. I hadn't played with it enough to notice posterization but I appreciate being alerted to this pitfall. I find channel blending to produce the best contrast based on the default 3-6-1 Photoshop does for R-G-B to be a bit haphazard with lots of trial and error...

any other thoughts?

Thanks so very much,

Bill Theis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Stephen Marsh
Wed Aug 8, 2007 2:53 am (PST)

William Theis wrote:

I think the technique has real promise and am surprised at the lack of
response!

I am not surprised Bill, when it comes to sharing methods and techniques the list is often quiet, while there will be much traffic on other areas of discussion that is not related to the application of theory. Yours is the first mention of this technique that I described in January that I have seen on any list Bill.

Glad you tipped me off to the other links, which I will
study. I hadn't played with it enough to notice posterization but I
appreciate being alerted to this pitfall.

I think this was with blue sky (not sure if the original was pristine or JPEG). It depends on the image and slider choice on whether posterization may appear, inspect at 100% zoom at least and look over the image carefully while toggling the adjustment preview on/off.

Regards,

Stephen Marsh.
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Wed Aug 8, 2007 7:12 am (PST)

Stephen Marsh writes,

Hi Bill, or one can switch to luminosity mode first and evaluate in
colour the luminosity blending effects as you move the sliders to add
drama to sky or separate foliage from backgrounds or whatever the task
may be.

When the B&W command was first announced, I thought it would have great potential in such a scenario (put it on an adjustment layer in luminosity mode, and adjust to taste). However, I was envisioning it as dividing the file into, in effect, a six-color separation (RGBCMY) and allowing each hypothetical channel to be worked with along the lines of Channel Mixer. That would have been helpful because the hypothetical "red" channel of such a document would be more precise in terms of what we might want to do, than the actual red channel of an RGB file.

Instead, its masks are color-based only: they identify areas that are "red", and are lighter or darker depending upon how pure the red is, not how dark or light the red is. The result is basically equivalent to using Selective Color as the adjustment layer, although the interface is slightly different. The magenta "channel", for example, is nonexistent unless Photoshop identifies an object somewhere as being magenta.

This is rarely the best way to improve luminosity, because so many objects are out of whack in relation to their actual darkness, and thus create contrast problems. For example, in a face, even of someone not wearing lipstick, the lips are somewhat darker than the rest of the face, which is irrelevant to this command. They are, however *much* redder than the rest, so they (or rather their most colorful parts) are affected disproportionately.

If we try to lighten the face by adjusting Reds or Yellows with the B&W command, the lips lighten faster than the rest of the skin, which looks bizarre. If we try to darken the face, not only do the lips get darker faster than the rest, but they lose detail: the darkest parts of the lips are less red than the lighter parts, so they darken less rapidly, thus the range closes and detail is lost. In either example, we'd be better off using an existing channel, probably the green. We could apply a curve to it in luminosity mode, blend it with other channels in luminosity mode, or use Channel Mixer in luminosity mode. Any of the three would give a better result than B&W or Selective Color.

I only played during the beta and not since. I found it interactive
and fast but easy to create posterization and artifacts with some
combinations of sliders and image colours/tones etc.

That's the kicker, and we should discuss why, because it helps understand some LAB problems also.

In principle, we should still be able to use this command to lighten or darken certain objects that don't contain much contrast themselves. In practice, the moves have to be very slight unless we go to considerable extra work.

For example, we often want skies to be darker and richer than found in the original photograph. We frequently use channel tricks of one kind or another to accomplish this. One would think that we might be able to use B&W on a luminosity layer also: we darken Blues (chances are there isn't anything else in the file that Photoshop identifies as being blue, and if there is we can layer-mask it out easily enough). In fact, you might think that this would be *better* than doing it the other way, because the darker the sky is, the bluer it is. So, darkening Blues using B/W or Selective Color in luminosity mode would accentuate the darker areas of the sky more than cloudy areas, which is probably exactly what we want.

The difficulty is the structure of the mask. In color correction, we often keep the concepts of color and contrast separate. It's useful to think of a camera in the same way. Cameras are much, much, better at evaluating contrast (lightness to darkness) than they are at evaluating color. Any camera is able to capture a pretty good grayscale image. But capturing delicate color variation *independently of darkness* is a different story.

Everybody knows that the blue channel of RGB is typically pretty choppy. This happens because the camera is so bad at evaluating blueness. However, this defect is largely hidden in the blue of RGB because that channel is also heavily based on darkness, which the camera captures accurately.

When Ric Cohn last year opened a thread on using the B channel of LAB as a mask, the question came up of why the B (which covers yellow/blue) should be so much less accurate than the blue of RGB, on which you might think it was based. The answer is that the defect is no longer hidden by what amounts to blending with a grayscale image--the B channel is color information only, and lightness/darkness has nothing to do with it. So it's very choppy, and if we want to increase its contrast and use it as a mask, it has to be heavily blurred.

The same thing applies to the quasi-channels used in Selective Color, Hue/Saturation, or B/W. As they are based on color only, independent of darkness, they're choppy and inaccurate. If we try to darken a sky by using the real red ch annel for a luminosity blend, noise can become a problem, as was pointed out in a recent thread by Ron Kelly. But if we try to do it by darkening the Reds in Selective Color or B&W, noise becomes an *enormous* problem--anything more than a nominal move causes the sky to break apart.

I've had some luck with extracting color-based masks and using them as layer masks in several different contexts. However, they always need further attention. For example, if we want to get the effect of a darker sky that accentuates the darker areas more than the clouds, here's a reasonable procedure.

1. Make a duplicate file.

2. Working with this duplicate file, make a duplicate layer.

3. Add a B&W adjustment layer. Exit it and set its mode to Luminosity.

4. Reopen the B&W command on the adjustment layer, choose Cyans (or possibly Blues, or both) and darken heavily. This should yield generally the desired effect, but nastily noisy.

5. Merge the adjustment layer with the middle layer.

6. Change mode of the newly merged top layer to Difference. Document now looks totally dark, except in areas that were affected by the darkening of the blues.

7. Flatten image.

8. Convert to grayscale.

9. Auto Levels.

10. Using the Surface Blur command, blur the noise out of the sky, while trying to retain distinction between clouds and sky. Typically a Radius of 10+ will be necessary.

11. Return to the original file. Make a duplicate layer.

12. Set the duplicate layer to Multiply mode.

13. Establish a layer mask. Load the duplicate file as the mask.

This should give a much smoother and more agreeable effect.

Dan Margulis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: "Rick Gordon"
Wed Aug 8, 2007 2:30 pm (PST)

This is where I often find a two-phase approach useful.

1) Do what you can do in RGB, i.e., the Black & White layer.

2) Stamp the resulting layer above (Cmd-Opt-Shift-E) in normal mode.

3) Delete or hide the Black & White Layer.

4) Convert to Lab.

5) Limit the results of the stamped layer with Blend-if sliders, based on a, b, and/or L values.

This general approach works for a wide range of issues, not just the Black & White adjustment.

Rick Gordon
___________________________________________________

RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________________

WWW: http://www.shelterpub.com
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Les De Moss
Wed Aug 8, 2007 5:11 pm (PST)

13. Establish a layer mask. Load the duplicate file as the mask.

I assume that portions of the mask affecting areas other than those targeted (the blue sky), are painted black, yes?

Les De Moss
DigiGraphics LLC
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Wai-hong Chung"
Thu Aug 9, 2007 4:54 am (PST)

Dan Margulis wrote:-

For example, if we want to get the effect of a darker sky that accentuates
the darker areas more than the clouds, here's a reasonable procedure.

1. Make a duplicate file.

2. Working with this duplicate file, make a duplicate layer.

3. Add a B&W adjustment layer. Exit it and set its mode to Luminosity.

4. Reopen the B&W command on the adjustment layer, choose Cyans (or possibly
Blues, or both) and darken heavily. This should yield generally the desired
effect, but nastily noisy.

5. Merge the adjustment layer with the middle layer.

6. Change mode of the newly merged top layer to Difference. Document now
looks totally dark, except in areas that were affected by the darkening of the
blues.

7. Flatten image

8. Convert to grayscale.

9. Auto Levels.

10. Using the Surface Blur command, blur the noise out of the sky, while
trying to retain distinction between clouds and sky. Typically a Radius of 10+
will be necessary.

11. Return to the original file. Make a duplicate layer.

12. Set the duplicate layer to Multiply mode.

13. Establish a layer mask. Load the duplicate file as the mask.

This should give a much smoother and more agreeable effect.

I just tried the above method and its fantastic! Its better than using channel blending. Thank you Dan for your advice!

Wai-hong Chung from Hong Kong
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Thu Aug 9, 2007 9:29 am (PST)

Les De Moss writes,

I assume that portions of the mask affecting areas other than those targeted
(the blue sky), are painted black, yes?

They already should be black. They wouldn't ahve been affected by the adjustment layer unless they already had a blue component. When Difference mode is used and there is no difference between the top and bottom layer, the result is black.

If certain parts of the image other than the sky are not quite black, then you can paint them in or lasso them out of the layer mask if you like, but you should take a book first, because the image may look more attractive if those areas are left alone in the layer mask.

Dan Margulis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: "Lee Varis"
Thu Aug 9, 2007 9:31 am (PST)

I've been experimenting with this myself - The tool seems to be based more on the old Hue/Saturation adjustment tool. Hue ranges are rotated, lightened or darkened using similar algorithms and, as such, are prone to the same sort of posterizing and banding artifacts as Hue/Sat is. It may seem more intuitive to use than Channel Mixer for B +W but it doesn't deliver as clean (or smooth) a result. While it is nice to be able to select a specific range of color to lighten or darken in grayscale you can't adjust the transition from one range to another the way you can with the old Hue/Sat command. You can achieve nearly identical results by using a Hue/Sat adjustment layer under some kind of de-saturation adjustment layer set to color like the
Russel Brown's method using two Hue/Sat layers.

regards,

Lee Varis

President, LADIG
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Les De Moss
Thu Aug 9, 2007 10:01 am (PST)

Thank you, Dan.

The image I experimented with was a mountain scene containing various shades of foliage in the foreground. Those areas did not go black in the mask, which made me think I needed to touch up the mask, or had not followed the instructions - always a distinct possibility. The results in the sky were effective and pleasing. Thanks for sharing this with the list.

Les De Moss
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Thu Aug 9, 2007 12:52 pm (PST)

I would guess that the mountains are somewhat hazy, which would give them a cyan component that it would share with the sky. I assume that you chose Cyans when working with the B&W command.

So, no, this wouldn't be a problem with your procedure. The only question would be whether you feel that since the sky got darker, the mountains should darken slightly as well. If yes, stick with the mask as is; if no, take it to black in the mountains.

Dan Margulis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: "williamtheis"
Tue Aug 14, 2007 10:58 pm (PST)

So you have convinced me that the CS3 B&W is not the preferred method for the reasons cited. However IMHO this does not negate the usefulness in visualization it offers: by viewing the image and using only the R,G,B sliders I think this could be used to tip me off as to which colors are the "friends" and "foes" to guide me in the final channel blend. After exploring the options, the B&W filter would be cancelled rather than applied. I know Dan would probably say this information could equally (or perhaps better) be obtained just by looking at the channels... which will come in my case with practice

Lastly there are a plethora of alternative B&W conversion processes, like the Russel Brown saturation layer techniques mentioned by Lee. Since Brown's "filter" layer is a hue layer (color mode), I would think it falls into the same pitfalls as the CS3 B&W or channel mixer. some of the other actions on the Adobe site substitue a curves, channel mixer, or other layer (color mode) below the desaturation layer. one of them converts to LAB and puts a L 100% then three layers with the, R, G, B channels on top with 0% opacity, which you then adjust to blend in a part of that color channel; the result then can be flattened and moved back to the RGB file

so if i am to understand this, the CS3 B&W filter "is color-based only: it identifies areas that are "red", and "are lighter or darker depending upon how pure the red is, not how dark or light the red is." if there were to be some kind of LAB analogue to perform B&W conversions, then the a and b which should also be based on how pure the red is (in this example), correct? all the lightness/darkness of said "red" would be in the L. a desaturation layer in LAB appears just to pick the L but you can apply a curves layer, flatten and copy the result back to the RGB and change to Luminosity blend...

Bill Theis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Mark Segal
Wed Aug 15, 2007 1:58 pm (PST)

One can go on and on playing with all the complex B&W conversion processes one wants, but in the final analysis from my experience I think it is largely a waste of time. I have a file-folder full of them from numerous sources going back over several years. What I rely on most for sheer ease of use, scope and flexibility to bring out all kinds of effects and quality of outcomes are (1) the greyscale option in the HSL menu of Camera Raw 4.x or a plug-in called "Convert to Black and White Pro" from "theimagingfactory.com". I think it is really worthwhile trying to exhaust the potential of these tools, then decide whether more exotic approaches would provide significant value-added.

Mark Segal
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Howard Smith
Wed Aug 15, 2007 3:14 pm (PST)

Bill, that was a well-thought out discussion. It puzzles me, though, that a single tool-if that's what it is (I'm still on CS) is not really necessary for converting images to B&W. There are so many quick, effective ways to do it that it's harder to choose among them than it is to get the job done.

Am I the only one who copies a likely channel and uses Curves to make a good B&W? Or is there something wrong with this that so far has not been apparent to me? It's extremely flexible, not to mention fast. Anyone who feels more comfortable doing it the hard way has Calculations available, but it takes longer and is too much like work.

On the other hand, maybe I shouldn't ask anyone to correct me. Maybe I'm better off remaining in ignorance. Not being inclined to be a theoretician, my thought is that if it works, it's good.

Howard Smith
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Thu Aug 16, 2007 3:46 pm (PST)

Bill Theis writes,

So you have convinced me that the CS3 B&W is not the preferred method
for the reasons cited. However IMHO this does not negate the
usefulness in visualization it offers: by viewing the image and
using only the R,G,B sliders I think this could be used to tip me off
as to which colors are the "friends" and "foes" to guide me in the
final channel blend. After exploring the options, the B&W filter
would be cancelled rather than applied. I know Dan would probably
say this information could equally (or perhaps better) be obtained
just by looking at the channels... which will come in my case with
practice

I don't think that this command offers a valid preview for this purpose. Usually, the channel that gets emphasized when blending to make a good B/W is neither lightest nor darkest in the object in which we're trying to gain contrast. but rather the channel in the middle. For example, in making a B/W of a face, normally the best channel is not the red (lightest) nor the blue (darkest) but
rather the green, which falls in the middle.

Unfortunately, that means that there is no color within the B/W command that can preview it. Faces are clearly neither green nor its direct opponent, magenta, so accessing either of these colors does nothing in the B/W command. We could darken the faces by using Oranges in B/W. but it wouldn't have nearly the contrast of blending with the green channel, in addition to adding unwanted noise.

Lastly there are a plethora of alternative B&W conversion processes,
like the Russel Brown saturation layer techniques mentioned by Lee.
Since Brown's "filter" layer is a hue layer (color mode), I would
think it falls into the same pitfalls as the CS3 B&W or channel
mixer. some of the other actions on the Adobe site substitue a
curves, channel mixer, or other layer (color mode) below the
desaturation layer. one of them converts to LAB and puts a L 100%
then three layers with the, R, G, B channels on top with 0% opacity,
which you then adjust to blend in a part of that color channel; the
result then can be flattened and moved back to the RGB file

There certainly are a lot of valid methods to make a good B/W. The key always is to find areas of color contrast and convert them into luminosity contrast. Methods that use color, rather than contrast, for the necessary blending run into noise problems fairly quickly.

so if i am to understand this, the CS3 B&W filter "is color-based
only: it identifies areas that are "red", and "are lighter or darker
depending upon how pure the red is, not how dark or light the red
is." if there were to be some kind of LAB analogue to perform B&W
conversions, then the a and b which should also be based on how pure
the red is (in this example), correct?

Blending directly with the A or B is dangerous for the same reason that the "channels" within the B/W command are--color channels are very noisy. You can use them in modest quantities, but if a big move is desired they need to be copied into a separate channel and blurred, then used as a mask, in the same way that I described earlier where the object was to intensify a sky using the Cyans "channel" of the B/W command as a mask.

Dan Margulis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Mark Segal
Fri Aug 17, 2007 3:43 am (PST)

I'm really baffled.I wonder whether Dan pulled-up some images and tested converting them to B&W using this new module in CS3 before posting the message below. I have, (even just now before I wrote this to make sure I believe what I'm saying) and I like it. Not quite as much as the greyscale conversion menu in CR 4.1, but nonetheless it is truly very good. It has individual controls for the rendition of luminosity in each of the six primaries (R, G, B, C, M,Y), and as one adjusts those controls one sees exactly what one is getting - all the more so with Soft Proof activated. Being an adjustment layer all settings are changeable. It produces gorgeous results with the same preview quality as any other soft-proof. Furthermore, it is then possible to add Curves or Levels Adjustment layers clipped to the B&W adjustment layer for producing additional contrast effects. There is truly a lot of flexibility here to produce stunning B&W images. Noise should be a non-issue by the time one gets to the B&W conversion stage. In a normal, logically ordered digital workflow one would implement noise reduction before Capture Sharpening. What's more, given the convenience of the interface, the time needed to explore numerous permutations and combinations of effects is minimal - just six sliders to move back and forth.

In support of these propositions, I refer readers to "Mastering Digital Black and White - A Photographer's Guide to High Quality Black-and-White Imaging and Printing" by Amadou Diallo, a noted expert in this field, published by Thompson Course Technology, Series Editor Harald Johnson, where Amadou reviews a slew of methods for doing B&W conversions from the simplest to the most intricate. Here is what he says about the CS3 Black and White command on pages 211 and 212: "The new Black and White command in CS3 is, by itself, a compelling reason to upgrade if you regularly convert color images to black and white." He then goes on to explain. Most importantly he says: "The brand new algorithms for these sliders do not require that you stick to combinations totaling 100 percent in order to prevent clipping......................Perhaps the most unique feature of this new tool is the ability to quickly identify the dominant hue at any given image location."

The new paradigm in Camera Raw and Photoshop is clearly to increasingly replace the need for convoluted workflows with much simpler, powerful techniques to achieve the same objectives. As I suggested the other day - it is good to explore these to the max, then decide on the value-added of more onerous alternatives.

Mark Segal
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Fri Aug 17, 2007 5:51 am (PST)

Howard writes,

Am I the only one who copies a likely channel and uses Curves to make a good
B&W? Or is there something wrong with this that so far has not been
apparent to me? It's extremely flexible, not to mention fast. Anyone who
feels more comfortable doing it the hard way has Calculations available, but
it takes longer and is too much like work.

No, it's the sensible approach for about a third or all images--those where one of the channels is clearly better than the other two in every respect. In that case, why not discard the garbage and proceed with the good channel?

The most common example is images of faces, where the green channel is usually clearly better than the other two.

As images get more complex, additional steps may be needed. For example, if the face is that of a woman wearing lipstick, the lips may turn out too dark if we use the green channel alone, and a curve may not be able to reach them. In that case, we first blend the blue into the green, Lighten mode, because the lips are likely to be the only area in which the blue channel is lighter than the green.

Dan Margulis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Stephen Marsh
Fri Aug 17, 2007 7:12 am (PST)

Mark Segal wrote:

One can go on and on playing with all the complex B&W conversion
processes one wants but in the final analysis from my experience I
think it is largely a waste of time. I have a file-folder full of them
from numerous sources going back over several years.

I think that I will Mark, it is a good way to learn, sometimes unexpected things and to reinforce other known points.

I gave up collecting and just tried to understand them, so that I could roll my own variation at will. I agree is some respects that it can be a waste of time, while from an educational perspective it is never a waste of time if one learns something.

What I rely on most for sheer ease of use, scope and flexibility to
bring out all kinds of effects and quality of outcomes are (1) the
greyscale option in the HSL menu of Camera Raw 4.x or a plug-in called
"Convert to Black and White Pro" from "theimagingfactory.com". I think
it is really worthwhile trying to exhaust the potential of these
tools, then decide whether more exotic approaches would provide
significant value-added.

My outlook has always been to exploit the investment in Photoshop or other native software to the full, before seeking other options. That being said, when other software outperforms a native Photoshop function and provides a competitive edge then I am all for it.

Sincerely,

Stephen Marsh.
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Howard Smith
Fri Aug 17, 2007 8:37 am (PST)

Thank you, Dan. I hadn't thought of channel blending with this approach. Another thing that seems to be helpful is to do a color and contrast correction before choosing a channel. In some images it makes for a significant improvement in the final result, especially for an image like Kodak's girl with the red hat.

Howard Smith
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Les De Moss
Fri Aug 17, 2007 11:59 am (PST)

Howard- Taking the single channel approach a little further, on occasion I have copied all three RGB channels to individual layers in a new grayscale document, applying curves, opacity and/or blending as necessary. This has the benefit of providing a grayscale preview of the changes to the image. This may be moot with CS3's BW conversion tool.

-Les De Moss ___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Fri Aug 17, 2007 9:11 pm (PST)

Mark Segal writes,

I'm really baffled. I wonder whether Dan pulled-up some images and tested
converting them to B&W using this new module in CS3 before posting the
message below.

The question I was responding to was not whether B&W is neat and/or keen, but whether it can be used to visualize the effect of blending one RGB channel into another. The answer is that it cannot, because a channel has extensive luminosity information whereas the B&W command works with color information only. There is no opprobrium imputed to B&W by saying this. The Unsharp Mask filter, for example, not to mention the History palette, do not do a good job of predicting channel blend results either, and yet they are valuable parts of Photoshop.

I have been attempting to respond to list messages in the order they were posted, including one by Mark about Camera Raw, but am breaking the sequence here due to a particularly urgent need described below.

In support of these propositions, I refer readers to "Mastering Digital Black
and White - A Photographer's Guide to High Quality Black-and-White Imaging
and Printing" by Amadou Diallo, a noted expert in this field,

Never heard of him.

Here is what he says about the CS3 Black and White command on pages
211 and 212: "The new Black and White command in CS3 is, by itself, a
compelling reason to upgrade if you regularly convert color images to black
and white."

Similar claims were made about the Smart Blur filter when it was introduced.

He then goes on to explain. Most importantly he says: "The brand new
algorithms for these sliders do not require that you stick to combinations
totaling 100 percent in order to prevent clipping......................Perhaps the most
unique feature of this new tool is the ability to quickly identify the dominant
hue at any given image location."

I do not understand the first half of the statement, except for "brand new algorithms", which these are not. As for the ability to help us identify that a stop sign is red, or that a tree is green, this is doubtless of considerable value to those color-blind individuals who do not know how to use the Info palette. However, the feature is not unique--Selective Color and Hue/Saturation do the same thing.

Returning now to what Mark said:

I wonder whether Dan pulled-up some images and tested converting them
to B&W using this new module in CS3 before posting the message below. I
have, (even just now before I wrote this to make sure I believe what I'm
saying) and I like it...It produces gorgeous results with the same preview
quality as any other soft-proof. Furthermore, it is then possible to add Curves
or Levels Adjustment layers clipped to the B&W adjustment layer for
producing additional contrast effects. There is truly a lot of flexibility here to
produce stunning B&W images.

As for testing, yes. Converting color files to B/W may not sound like a particularly interesting activity but in fact it is of critical importance in color correction, because it is the foundation of luminosity-based moves involving curves and, particularly, channel blending. List members are well aware of the significance of this topic; increased coverage of blending was the biggest request for PP5E. It continues to be a topic of great interest to the list. At Photoshop World/Boston, hundreds of people were turned away from my session on blending and even so, so many people were seated on the floor that the fire marshal had to order them to leave.

Consequently, any improvement in this procedure would be a very big deal indeed. That's why, when I first saw the written description of what was new in Photoshop CS3, I immediately put this new command on the top of the list of things to be checked out because of their potential workflow implications, and I said so publicly. Stephen Marsh said the same thing before the release, based only on the written description. When I first was exposed to the shipping version of CS3, please be assured that this was the very first command I went to, because if it had been designed as I imagined, it would have had a lot of impact.

Instead, it's just a rehash of existing commands--nothing of importance that Selective Color can't do just as well. But of course, nobody makes B/Ws using Selective Color, because it's not an effective way. I've explained why I think that in message 18357, which also emphasizes that I took the announcement of the command seriously. However, as I indicated, I think it has little value in creating B/Ws, although there are some mask-generation possibilities, one of which I illustrated to the list.

Now, it appears that you have found that "stunning B&W images" can be produced from this command. If so, I am not able to reproduce this result. Fortunately, there is a way to verify whether it is true. The topic is so important that each year I do a section on it in my advanced courses--a two-hour lecture followed by a five-image class exercise.

For this year, I have eight images that I am considering as useful examples. I do not have permission to post several of them, but in at least four cases I * can* post the original color file, if you are willing to demonstrate how you would produce your stunning results. In some of these cases I also have some of the better results from students that we could use for comparison to your work. No fair looking at their results before you start; they didn't get a chance to look at yours.

The rules of engagement are that selections and masks, including layer masks, may not be used. Also, it is prohibited to use painting tools or anything else found in the toolbox. All other methods are kosher. The final result is to be viewed with grayscale set to 20% dot gain.

If you are able to show competitive results, I would very much like to show my students how you got them. Also, it's of particularly great importance for Photoshop World, as described below.

The new paradigm in Camera Raw and Photoshop is clearly to increasingly
replace the need for convoluted workflows with much simpler, powerful
techniques to achieve the same objectives. As I suggested the other day - it is
good to explore these to the max, then decide on the value-added of more
onerous alternatives.

In the workflow referred to by George Machan, there is in fact a convert-to-B/W step. Because the *entire* correction must be accomplished in 3-5 minutes, there is no time for anything "onerous": for this step I can spare 20-30 seconds, not more. If you are able to demonstrate a superiority (or even something competitive) for using the B&W command, I would certainly change my presentation and I know that many of the attendees would be grateful. However, I would need to see the corrected images early next week, as beyond that time I will be beyond the reach of the web. That is why I am replying to this message out of order.

Please let us know if you are willing to assist these attendees in this fashion. If so, I will be happy to post the base files.

Dan Margulis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Richard Wagner
Sat Aug 18, 2007 7:40 am (PST)

On Aug 17, 2007, at 8:55 PM, Dan Margulis wrote:

I do not understand the first half of the statement, except for
"brand new algorithms", which these are not.

So, you can state for a fact that there are no new algorithms underlying the B&W command? How is it that you know this for a fact? Can you really test all possible inputs so that you can see if they match your predictions based on the "old algorithms?" As a programmer, it just baffles me that you are able to make this assertion without inside information from the programming team, but rather, you are able to deduce this in some manner. I'd really like to know how.

--Rich Wagner
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Mark Segal
Sat Aug 18, 2007 10:36 pm (PST)

Hi Dan,

True - you can't see channel information from that filter, because once you use the filter you are not supposed to have to worry about blending channels - that's the principle. so if the context is to learn about luminosity from channels, this won't help, but I thought the context is making B&W photographs. More about that below.

As you haven't heard of Amadou Diallo, FYI, he is a New York City based photographer, writer, digital imaging consultant and teacher. He owns and operates "Diallo Photography", a studio that offers fine art printmaking and digital imaging workshops to an international clientele. You can see more about him at www.diallophotography.com. By the way, as I mentioned, the Series Editor is Harald Johnson; you most likely know of Harald, a noted international authority on digital printing.

Now turning to the substance of B&W conversion in Photoshop, I'd like to ask how you know the algorithms are not new? And in the same vein I guess one could also ask Amadou how he knows they are. One thing for sure: the tool is certainly new to Photoshop. I don't think "Selective Color" and "Hue Saturation" are relevant to what he was saying in the context of producing greyscale images. It was quite clear to me what Amadou meant by that statement, given the context, is that it allows one to easily identify the impact of particular hues on the greyscale rendition of those hues. "Selective Color" was not primarily intended for the conversion of RGB to greyscale, whereas the B&W tool is. Whether or not the B&W Adjustment Layer is a "rehash of existing commands" is neither here nor there - I expect that no-one other than the programmers themselves would know this. But the important thing - no matter how it was configured - is whether it works well.

Now, you talk about channel blending as "a procedure" - OK generically so, but as you know better than me, there are all kinds of channel blending one can do depending on the purpose, and there are all kinds of purposes. The only purpose I'm writing about is to make a high quality black and white photograph from an RGB original file, not per se to reveal new insights about the luminosity information contained in channels.

Hence, my only interest in this is whether indeed using the B&W Adjustment Layer can produce as good a B&W photograph as can be obtained from more complicated techniques. If you and I agree on this "objective function", I'm willing to play the game. You talk about three to five minutes. I can live with that. The quick job I did the other night fits easily within that time frame from acquisition of the raw file into Camera Raw until ready for printing. Your other "rules of engagement" are fine, except for the last point which raises a question about the destination of the file. My files are purposed to an Epson 4800 inkjet printer using Epson Enhanced Matte paper, hence I evaluate my images under softproof in Photoshop using the Epson Pro4800_EMK profile, and then of course I print them, because that's the bottom line - but the softproof and print are usually coherent because my colour management set-up is satisfactory. Dot gain is not relevant to this workflow, so any consideration of effects from dot gain is outside my frame of evaluation. That said I recognize its relevance for pre-press, but I make this point just in case the outcome of a comparison of techniques can be affected by the purposing of the file.

There is however some procedure we need to discuss. I'm intrigued by the idea of comparing the outcomes of what Photoshop proposes by way of the "one-stop" approach (the new B&W Adjustment Layer) versus what Photoshop facilitates using the alternative workflow you and others have developed using channel blending. As you say, this could be of significant educational interest. However, to do that properly, the same image must be used for comparison of the alternative techniques. So the way I would think sensible to set this up is that subject to an agreement between us on the procedure, I shall provide you with both the raw file and the work-up I did of the same image referenced in this discussion. I'll send you a jpg of it in a private email just after I send this one, since nothing can be attached to list messages. We will need access to a server (and the FTP address) for exchanging files, because they are large and they would need to remain unflattened along with the history data to preserve transparency of workflow on all sides. I work in 16-bit ProPhoto space, and I would not conduct these exercises with JPEGs. My raw files are typically in the range of 9~10 MB and the processed version of this file is 117 MB unflattened. I'll also send you the Epson Pro4800_MK printer profile so that you can install it in your system for one version of soft-proofing your own results. (I would need to see the comparison using this soft-proof, whereas probably the CMYK at 20% dot gain result would be more relevant to your purposes - both are interesting and important to the clientele.) Once you have my file and examine what I did and how I did it, you would try to improve upon it starting from the raw file and ending-up with the final PSD file using your own techniques, which file you would post to the server unflattened with the history preserved - I would want the one purposed for the Epson 4800 softproof because I know those images. Once I see your result, I would then want to see whether by any further tweaking within the "kosher" rules I can simulate what you got, as the results are likely to be different (whether one is "better" than the other depends on the criteria for defining "better", but that is something we can discuss anon.)

Once we've been through this exchange of technical approaches, if you want to use the image for showing both your results and mine in your classes and at PhotoshopWorld, I have no objection provided the authorship of the image is acknowledged.

Mark Segal
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: "Rick Gordon"
Sat Aug 18, 2007 10:46 pm (PST)

I would love to have a crack at the files. I also think that this same protocol would do well to move the "Camera Raw and Range Setting" discussion into something more concrete.
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: dave_cardinal
Sun Aug 19, 2007 10:49 pm (PST)

Since I learn a lot from reading this list, I'd be delighted to provide an ftp site for images which are illustrative on this point
or others discussed on the list if it'd be helpful.

--David Cardinal
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Mon Aug 20, 2007 2:47 pm (PST)

Mark Segal writes.

By the way, as I mentioned, the Series Editor is Harald Johnson; you most
likely know of Harald, a noted international authority on digital printing.

Indeed I do.

Now turning to the substance of B&W conversion in Photoshop, I'd like to
ask how you know the algorithms are not new?

I scripted a method of depositing the data into a spread--oops, pardon me. Forgot.

Since you inquire, I was informed of this fact in a vision. The Archangel Gabriel appeared, and revealed it to me in response to a prayer. But as he dematerialized, he was impertinent enough to whisper in my ear, "You know, you really could have figured this out without my help."

There is however some procedure we need to discuss. I'm intrigued by the
idea of comparing the outcomes of what Photoshop proposes by way of the
"one-stop" approach (the new B&W Adjustment Layer) versus what
Photoshop facilitates using the alternative workflow you and others have
developed using channel blending. As you say, this could be of significant
educational interest. However, to do that properly, the same image must be
used for comparison of the alternative techniques. So the way I would think
sensible to set this up is that subject to an agreement between us on the
procedure, I shall provide you with both the raw file and the work-up I did of
the same image referenced in this discussion.

Somehow I had a feeling that this or something similar would be the reply. Thanks for the offer, but I think I'll settle for whatever techniques people bring to my advanced classes later this year. If you're in Toronto that week, by all means stop in for the session--you'd be the only Canadian in the room.

Dan Margulis
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Mark Segal
Mon Aug 20, 2007 8:41 pm (PST)

OK Dan, you stopped half-way so I WON'T ask for the spread//............:-).

I much appreciate your invite into that session of your Toronto class and I shall attend. Please send me the date off-List. Notwithstanding your response on the image, I think it would really be neat for your students in that class - amongst whatever else they do - to play with this one image I worked-up the other day, I truly am intellectually curious to see whether they can better what I can do with the B&W Adjustment Layer on that image. Who knows, depending on what comes out of that comparison, there may be an interesting story to tell people - but what that story would be we won't know until several people devote their skills to working it up. If you wish to reconsider and I should send it to you please let me know off-List.

Mark Segal
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: "Howie"
Mon Sep 10, 2007 1:29 pm (PST)

Dan Margulis wrote:

However, I was envisioning it as dividing the file into, in
effect, a six-color separation (RGBCMY) and allowing each hypothetical
channel to be worked with along the lines of Channel Mixer. That would have been
helpful because the hypothetical "red" channel of such a document would be more
precise in terms of what we might want to do, than the actual red channel of an
RGB file.

Why would such a channel be more precise?

Instead, its masks are color-based only: they identify areas that are "red",
and are lighter or darker depending upon how pure the red is, not how dark or
light the red is.

Is color purity defined by the range of hue found in a color, the degree of saturation, or some other way?

--
On a divergent note: Am I wrong in my understanding that this thread started with discussing the B&W command for adjusting contrast in color images and later became a dicussion on actual B&W conversion?

/howie nordström
howienordstrom.com
___________________________________________________________________________

Re: CS3's new B&W tool used for Contrast in color images
Posted by: Dan Margulis
Tue Sep 11, 2007 10:01 am (PST)

Howie writes,

Why would such a channel be more precise?

Before converting color>B/W the key is to identify areas where we are perceiving contrast as a result of variation in color, which won't survive the conversion, as opposed to luminosity, which will. Then, we have to figure out how to transform that color variation into luminosity variation.

The most effective way of doing this is to find appropriate detail in one of the existing channels and emphasize it in some manner, through curves or one of several blending methods. The detail always is found in colored areas, as in neutrals all channels are equal.

The more channels the file is divided into, the more likely we will be to be able to isolate this detail, since the more channels, the lighter the neutrals will be in each--easy to avoid.

In principle, CMYK ought to be better for such maneuvers than RGB, because of the extra channel. Unfortunately, the ink limit in CMYK hoses shadow areas, plus values for neutrals aren't equal in each channel, which is inconvenient for blending. So, RGB is as a practical matter a better choice, although many retouchers make a separate CMYK document and use the black channel for various blends.

We should be able to get more effective results if the file is split into artificial channels, say eight primary colors plus a black. When I first read the description of the new command, that's what I thought it might be.

Is color purity defined by the range of hue found in a color, the
degree of saturation, or some other way?

The command creates a mask based on hue, modified by saturation. Luminosity is not a factor, which makes it problematic for regaining detail.

On a divergent note: Am I wrong in my understanding that this thread
started with discussing the B&W command for adjusting contrast in
color images and later became a dicussion on actual B&W conversion?

For those participating in the conversation, there's little difference. We were all, I believe, thinking in terms of using this command on a layer set to Luminosity mode, with the objective of improving contrast in color images. But if it worked for that, it would work for creating a grayscale file.
I want to point out a useful new feature in CS3 for this type of blending. Two new modes are now available both in layers and in Apply Image and they usually (but not always) give better results than previous alternatives.

We often use Darken mode, which (using layers as an example, although it works the same way in other commands) combines each channel of each layer by comparing each pixel and choosing the darker one. IOW, it's possible to have a Darken layer that does nothing if it is everywhere lighter than what's underneath. No part of any channel can ever get lighter than it is on the bottom layer. But some channels can get darker while others are unchanged.

The new Darker Color mode eliminates this possibility. It compares the * overall* darknesses in each area, not channel by channel as Darken does. It replaces the bottom layer in areas where it concludes that the top layer is darker overall. It either replaces all channels at once, or none.

Example: on the top layer is a person's face, on the bottom a bright yellow flower. Overall, the face is a darker object. Even though the flower is lighter overall, its blue channel is much darker than the face's. If we set layer mode to Darker Color, the face takes precedence, period. If we set it to Darken, we get the red and green channels of the face married to the blue of the flower. Sometimes that's what we want, but we're more likely to want the Darker Color effect, particularly in retouching (it works with painting modes, too).

There is a similar relation between the existing Lighten and the new Lighter Color mode.

Dan Margulis