Dan Margulis Applied Color Theory

Gamma 1.0, Blurs, and Blends

Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: dale97305
Wed Nov 22, 2006 11:11 am (PST)

Dan mentions the option to "Blend RGB Colors Using Gamma 1.00" in Photoshop's Advanced Mode for color settings on pp101-103 of the LAB Color book. The illustrations 5.16 B & C make it very clear that there is a significant difference when pure RGB colors are blended.

Since I mainly use sRGB & Adobe RGB color spaces in processing my photos, I am wondering if this setting would be a good one to use as a default. I generally work with layers & masks to acomplish most of my image manipulations... tone, color, sharpening operations; etc.

Any pros or cons for making this a default for my regular processing workflow?

Dale Price
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: Stephen Marsh
Wed Nov 22, 2006 5:55 pm (PST)

Dale Price wrote:

Dan mentions the option to "Blend RGB Colors Using Gamma 1.00" in
Photoshop's Advanced Mode for color settings on pp101-103 of the LAB
Color book. The illustrations 5.16 B & C make it very clear that there
is a significant difference when pure RGB colors are blended.

Since I mainly use sRGB & Adobe RGB color spaces in processing my
photos, I am wondering if this setting would be a good one to use as a
default. I generally work with layers & masks to acomplish most of my
image manipulations... tone, color, sharpening operations; etc.

Any pros or cons for making this a default for my regular processing
workflow?

Dale, I will leave Dan to comment on how this relates to his book and suggested workflows- but it really depends on the working space being used and if you like the results as to whether you think this is a good thing. Do you like the way that Photoshop works by default when compared to a gamma 1 in colour settings?

It might be good for some edits where blending occurs, but perhaps not for other work (so is probably best left for special cases and not as a permanent colour settings default).

The image processing theory goes that all such moves should be done in a linear gamma space for a 'colorimetrically correct' result (which may not be what a *human* observer feels as correct or wanted, or it may!).

Some software uses gamma 1 in 32 bpc behind the scenes, even when working on say an sRGB image (the transforms are hidden).

I would suggest that you read this post of mine from 2002, where my initial findings after being made aware of this were that gamma 1.8 or that in CMYK or LAB provides a better blending for my personal taste over using a gamma of 1.00 or 2.2:

http: //tech.groups.yahoo.com/group/colortheory/message/4208

http: //www.aim-dtp.net/aim/photoshop/v6/blending/rgb_blending.htm

(beware the above long links breaking)

This is of course related to but not the same as working in the AIM-RGB colour space in Photoshop (gamma 1) - which has been trashed on many websites over the years...but now we find that ACR and Lightroom use a gamma 1 too, with the difference being that camera raw data is being used and not a processed image so perhaps this has more validity in a raw converter where the sensor records light in a linear fashion, but perhaps it does not make sense in Photoshop proper (as I have not seen anybody eat Timo's humble pie).

Hope this helps,

Stephen Marsh.
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: "Lee Varis"
Thu Nov 23, 2006 4:54 am (PST)

On Nov 22, 2006, at 10:49 AM, dale97305 wrote:

Any pros or cons for making this a default for my regular processing
workflow?

The biggest danger in using this as a default occurs when you share layered files with others. This may never happen for you but in the entertainment industry work that I do many people require layered Photoshop files that end up being shared with several collaborators. I encountered a situation where someone had a multi-layer composite that used some blending modes for special effects – they had achieved a certain look by utilizing the "Blend using gamma 1" setting. Of course this same layered file looked completely different on the next users computer because they were not using that default in their preferences! After several back and forth exchanges everything got very confused. I was involved in troubleshooting the problem which they thought was a color management issue! It took a long time of poking around until I figured out what was going on. Neither user was aware that their preference settings were different – they weren't supposed to be. This is not to say that this setting doesn't have any advantages – you just have to be aware of it when you share files.

You might never encounter a problem like this if you never send layered files to anyone.

regards,

Lee Varis

President, LADIG
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: Stephen Marsh
Thu Nov 23, 2006 6:06 am (PST)

Lee Varis wrote:

It took a long time of
poking around until I figured out what was going on. Neither user was
aware that their preference settings were different – they weren't
supposed to be. This is not to say that this setting doesn't have any
advantages – you just have to be aware of it when you share files.

You might never encounter a problem like this if you never send
layered files to anyone.

This is a good case (there are likely more) for saving and sharing the .csf colour settings file if in a workgroup exchanging master files.

Stephen Marsh.
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: "Michael King"
Thu Nov 23, 2006 10:43 am (PST)

So one thing I don't understand is, if gamma 1.0 ProPhoto (I know its a custom version) is good enough for Lightroom, why we are not all using something similar in Photoshop as well? I.E why don't we use a 1.0 gamma space for ALL editing in Photoshop? I am guessing there must be some reason for this, or is it just historical that we are still using working space gammas other than 1?

Timo seems to advocate this gamma 1.0 approach on AIM http://www.aim-dtp.net/aim/techniques/linear_raw/index.htm

Mike King
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: Dan Margulis
Thu Nov 23, 2006 10:43 am (PST)

Dale writes,

Dan mentions the option to "Blend RGB Colors Using Gamma 1.00" in
Photoshop's Advanced Mode for color settings on pp101-103 of the LAB
Color book. The illustrations 5.16 B & C make it very clear that there
is a significant difference when pure RGB colors are blended...
Any pros or cons for making this a default for my regular processing
workflow?

I don't make a recommendation. Checking it doesn't override every blending decision, so certain color blends won't look like certain others, which is trivial. It is at variance with traditional Photoshop practice, as Lee mentioned, so there may be a problem in interchanging files with others. Also, no other programs AFAIK make use of this feature, so if you blend colors in, say, InDesign, you'd get a different result than in Photoshop.

If none of these cons bother you, then I suppose checking it may be right, since the blends will certainly appear more accurate. OTOH, they still won't look as good as if done in LAB, so if faithful blends are a big priority, you probably wouldn't be working in RGB anyway.

Dan Margulis
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: “Andrew Rodney”
Thu Nov 23, 2006 12:01 pm (PST)

Raw files are linearly encoded, rendered files are not (they are usually a 1.8 or better 2.2 encoded). Have you ever seen a histogram of a linear file? Most of the tones are all shoved up to the left side. Also, with a raw file (or a linearly encoded file), the first half of all the data in the file defines the first stop of data found in the highlights. These sensors are simply photon counters and record the light striking them in a linear fashion. Each level recorded produces the same, corresponding level of data in the RAW file. If one photon strikes the sensor representing the first dark tone that can be recorded, the result is level 1 in a digital file. If 4096 photons strike the sensor, representing the brightest value it can record, then the result is level 4096 in a digital file. Fine since this is how a sensor records the data and fine for processing the data in a converter. Not so fine or easy once the image is rendered. In Lightroom and ACR, the histograms are based on a gamma corrected (sRGB TRC) encoding so that it matches what you get when you encoded in a gamma corrected working space.

We ideally like to work in a perceptually uniform working space. That means, equal amounts of a correction produce roughly an equal visual change on screen (which isn't a linear device by a long shot and almost always exhibiting a 2.0 to 2.2 TRC). In a linear encoded working space, if you were doing tone corrections, depending on where in the tone you were doing your adjustments, you'd have to move the sliders a great deal to see a small tweak or if you moved the slider a tiny amount, the effect would appear huge. Most users don't like that behavior. With a perceptually uniform working space, as you move the sliders and such, the edit appears smooth and uniform.

As for Timo, he's considered a bit of a wack job by just about any software engineer or color scientist who's ever made the mistake of debating him.

Andrew Rodney
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: "Iliah Borg"
Thu Nov 23, 2006 3:24 pm (PST)

Dear Andrew,

Thursday, November 23, 2006, 1:58:55 PM, you wrote:

Have you ever seen a histogram of a linear file?

Here is an image of normalised raw file with g=1 profile embedded:
http://www.pochtar.com/GMBCCDC_Linear.jpg

Most of the tones are all shoved up to the left side. Also, with a raw file
(or a linearly encoded file), the first half of all the data in the file
defines the first stop of data found in the highlights.

It is "first stop" of ADC, not just a "first stop" which often is understood as the first stop of DR in eV metrics.

These sensors are
simply photon counters and record the light striking them in a linear
fashion. Each level recorded produces the same, corresponding level of data
in the RAW file. If one photon strikes the sensor representing the first
dark tone that can be recorded, the result is level 1 in a digital file.

Not exactly. The result will be zero level. The ratio is about 1:12.

If 4096 photons strike the sensor, representing the brightest value it can
record, then the result is level 4096 in a digital file.

Similar, more like 320.

--
Best regards,
Iliah Borg
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: Richard Wagner
Thu Nov 23, 2006 3:40 pm (PST)

Michael,

Here is the long, but easy-to-read version of what Andrew described, including pictures of the histograms.

http://www.adobe.com/digitalimag/pdfs/linear_gamma.pdf

--Rich Wagner
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: "Michael King"
Thu Nov 23, 2006 5:21 pm (PST)

Richard and Andrew,

Many thanks for your info.

Mike
___________________________________________________________________________

Re: Using Gamma 1.00 setting for color blending in Photoshop CS
Posted by: Stephen Marsh
Thu Nov 23, 2006 5:31 pm (PST)

I am guessing there must be some reason
for this, or is it just historical that we are still using working space
gammas other than 1?

Mike, there is a good reason, do a search of this list or the ColorSync Users list it has been mentioned before why extreme gamma eiditing spaces are not good for the majority of images (but that is not to say that a polarbear in a snowstorm may be helped by using such a space).

As the advanced gamma blending box does not affect the blending of colours when blurring, one must use the gamma 1 edit space if wishing to have large averaging blurs working correctly (as in a 2.2 space they go grey, but in 1.8, CMYK or LAB they do not go gray).
 
So this is not be confused with the advanced checkbox for blending, but it is related to the same issue.

Best,

Stephen Marsh.
___________________________________________________________________________

Re[2]: [colortheory] Re: Using Gamma 1.00 setting for color blending
Posted by: "Iliah Borg"
Thu Nov 23, 2006 5:38 pm (PST)

Here is the long, but easy-to-read version of what Andrew described,
including pictures of the histograms.

http://www.adobe.com/digitalimag/pdfs/linear_gamma.pdf

Which is not very accurate, and becomes close to misleading when it comes to "a raw histogram would be a rather strange-looking beast, with all the data clumped at the shadow end," because it is presumed that no data normalisation was applied.

Looking into 12-bit data in 15-bit space without normalisation... It is the same as to say that 8-bit images are dark in 15 (they call it 16) bit space... Why overstate the matter?

--
Best regards,
Iliah Borg
___________________________________________________________________________

Re: Gamma and Blurs (was Using Gamma 1.00 setting)
Posted by: Dan Margulis
Fri Nov 24, 2006 10:05 am (PST)

So one thing I don't understand is, if gamma 1.0 ProPhoto (I know its a custom version) is good enough for Lightroom, why we are not all using
something similar in Photoshop as well? I.E why don't we use a 1.0 gamma
space for ALL editing in Photoshop?

Editing in very low gammas has some advantages but many disadvantages as well. The same is true of the L channel of LAB--it has (in effect) a very high gamma. Some things work better in it, others worse.

The reasons not to use it as a working space are general--most commands behave in unfamiliar ways, the files are likely to be misinterpreted at some point down the line, and perhaps worst of all, we can't evaluate the quality of the channels, because Photoshop unfortunately doesn't correct for gamma when it displays them. So, if we're editing at a very low gamma, Photoshop displays all three channels almost as black.

In working on PP5E, I found a use for low gamma that I had previously underestimated, so let me share it now.

Digicams, as we all know, are susceptible to unpleasant noise in the shadows, particularly if the image is underexposed or shot rapidly. Minimizing it by blurring can be problematic because the blur tends to affect important areas of the image too.

It turns out that this blurring works better in low gamma. Many people have made themselves false profiles of, say, Adobe RGB primaries with 1.0 gamma. The usual purpose is to assign this profile to some overly dark image and then convert it to a more customary RGB or to LAB, thus lightening the image without having to edit it.

To remove shadow noise effectively, we can *convert* the noisy image to this false profile, not assign it. This works because low-gamma spaces devote much less range to the shadows, and more to the light half of the image.

In the deepest shadows, serious noise may be represented by only a 2- or 3-level difference from its background if we are in low-gamma, but the same noise may be an 8- or 10-level difference in, say, Adobe RGB, whose gamma is 2.2.

A blur mild enough to reduce the variation by one level in the shadows of the low-gamma version has a big effect on the shadow noise but a much less noticeable effect elsewhere in the image.

The impact of the low gamma is especially pronounced when using the Surface Blur filter (CS2). This filter works like Gaussian Blur, except it tries to recognize edges and avoid blurring them. Its Threshold setting defines how big of a difference constitutes an edge; the higher the Threshold, the more objects get blurred. So, the objective is to keep the Threshold low enough to blur the areas you want without damaging anything else. (Warning: this filter is so heavyhanded that you want to keep it on a separate layer and reduce opacity).

In low gamma, a Threshold of 2 or 3 may substantially reduce shadow noise without affecting much else, whereas in normal gamma a value of 8 or 10 may be needed, and that will hose a lot of other things.

In short, when reducing shadow noise is a major goal, there's a good case for converting to low-gamma, doing the blurring, and then converting back.

Dan Margulis