Dan Margulis Applied Color Theory

Changes in Brightness in Preview


Changes in brightness in convert preview
Posted by: Ric Cohn
Sun Sep 16, 2007 9:23 pm (PST)

I'm wondering how much to trust the preview for different rendering intents. When using Convert To Profile when no colors are out of gamut, Relative Mode, as expected, shows virtually no change in preview appearance. However, for the icc profiles I've looked at, perceptual shows a change in brightness for the entire image- even the neutral and near neutral areas.

I've just been preparing some images for a publication that told me their printer uses SWOP2006_Coated3V2. When comparing Relative and Perceptual conversion preview, even when no colors are out of gamut, the image previews *lighter* in Perceptual Mode. However, when looking at the preview when converting the same image to US Web Coated (SWOP) v2 (or US Sheetfed Coated v2) the image previews *darker* in Perceptual Mode.

My questions for the group:
1. Is this accurate?
2. Does this reflect a change in preferences in building Perceptual Intent icc profiles?
3. Why do these 2 profiles appear to handle a Perceptual change so differently?
4. Does this difference suggest any changes in workflow when converting images?

Looking only at the preview, it makes the most sense to me to bring as many colors as I can into gamut and *always* use Relative. If important colors die when I try to bring them into gamut I'll leave them alone and try to improve the interpretation myself in CMYK. Is this a reasonable workflow? Are there times when it's better to use Perceptual?

Ric Cohn
___________________________________________________________________________

Re: Changes in brightness in convert preview
Posted by: Stephen Marsh
Mon Sep 17, 2007 6:47 am (PST)

Ric Cohn wrote:

I'm wondering how much to trust the preview for different rendering
intents.

The folk that build many profiles often talk about there being a different preview section of the profile, that can be tweaked independently of the conversion side of the profile, if the preview is found wanting but the seps are not. But I am confident that this is not what you are referring to Ric, I have mentioned this issue many times in the past on the list.

When using Convert To Profile when no colors are out of
gamut, Relative Mode, as expected, shows virtually no change in
preview appearance.

Agreed.

However, for the icc profiles I've looked at,
perceptual shows a change in brightness for the entire image- even
the neutral and near neutral areas.

Let me guess, the profile creation package is not Adobe, it is from GMB or somebody else known for profile creation software, a commercial ICC profile creation package (various free ICC profile inspection software tools will list this on Mac or Win).

I've just been preparing some images for a publication that told me
their printer uses SWOP2006_Coated3V2. When comparing Relative and
Perceptual conversion preview, even when no colors are out of gamut,
the image previews *lighter* in Perceptual Mode. However, when
looking at the preview when converting the same image to US Web
Coated (SWOP) v2 (or US Sheetfed Coated v2) the image previews
*darker* in Perceptual Mode.

Do the final CMYK numbers reflect the preview difference between the profiles and render intents? They should, this is not just preview.

As for the Adobe v2 SWOP (TR001), that would be slightly different from the revised 2006 version, would it not (is it still TR001)? Are you really comparing the same thing? How is the relcol render, similar to Adobe's relcol for SWOP in overall brightness?

Then there is an old debate going on here about the Adobe v2 SWOP profile and the flatsheet profile having TVI values that are unexpected by Dan, while some others have no problem with the dot gain results and can explain them to their own satisfaction.

My questions for the group:
1. Is this accurate?

I am not sure if "accurate" is a term to use for perceptual transforms. Will the result be pleasing, will it not be rejected? Are neutrals still neutral? Is the TAC honoured?

Is this behaviour expected of perceptual renders in non Adobe profiles? Yes.

Is it desired? Perhaps ask various profile builders and the users.

2. Does this reflect a change in preferences in building Perceptual
Intent icc profiles?

It reflects the profile package and settings built into the profile, by the person who made it, as far as I know (but it would also be good to know how much room there is for this tweak, if the person building the profile can back off or if they are stuck with what the profile package puts in as secret sauce). It is another option, how good an option depends on the person that built the profile and the software.

This is not a new thing, the Chromix TR001 family of GCR/UCR/TAC profiles have this secret sauce "kick" in them too, and they are as old as the v2 Adobe profiles which do not have this feature. Both describe the same condition almost exactly, but provide very different perceptual renders and to a lesser extent, different relative colorimetric renders.

3. Why do these 2 profiles appear to handle a Perceptual change so
differently?

The ICC does not specify exactly how a perceptual transform is to take place, only that it is one of the current four rendering possibilities, AFAIK. Adobe built their own profile creation software and their supplied profiles are very conservative in the perceptual table, it is closer to RelCol in results while still providing minor gamut compression (when compared to off the shelf profile packages that are more extreme). I give credit to Adobe for offering conservative profiles, as one can find less conservative ones for the same condition - but never as conservative as the Adobe profiles perceptual transform.

4. Does this difference suggest any changes in workflow when
converting images?

It may, depending on the results at proofing or press if nobody likes the result or if things are too light or whatever (it is probably best to do a trial proof or even better, print - if one can take the risk).

Are there times when it's better to use Perceptual?

Depends on the image or profile, I like to have the option, but that is all it is, another option to get me where I need to go...but yes, I have used perceptual a few times for really out of gamut detailed work, but generally RelCol is my mainstay. One can also blend the two together or mask etc.

For me it comes down to the numbers and the appearance of the file when sofproofing/proofing to a known condition, if in doubt I would be more conservative than not for coated web/flatsheet (newsprint I may be more aggressive and less conservative).

Some general info from a past thread here:

http: //tech.groups.yahoo.com/group/colortheory/message/278

I think it comes down to "secret sauce". Perhaps Terry or Chris or one of the other members known for commenting on ICC profile building would care to comment (or perhaps a lurker will delurk! Please!).

You could try reading this PDF, but I did not find it that useful (it only raised more questions that I am still in search of answers for):

http: //www.color.org/ICC_white_paper_1_Perceptual_rendering_fundamentals.pdf

Hope this helps,

Stephen Marsh.
___________________________________________________________________________

Re: Changes in brightness in convert preview
Posted by: Dan Margulis
Mon Oct 8, 2007 4:48 pm (PST)

Ric Cohn writes,

I've just been preparing some images for a publication that told me
their printer uses SWOP2006_Coated3V2. When comparing Relative and
Perceptual conversion preview, even when no colors are out of gamut,
the image previews *lighter* in Perceptual Mode. However, when
looking at the preview when converting the same image to US Web
Coated (SWOP) v2 (or US Sheetfed Coated v2) the image previews
*darker* in Perceptual Mode.

My questions for the group:
1. Is this accurate?

I don't see why the preview should be wrong.

2. Does this reflect a change in preferences in building Perceptual
Intent icc profiles?

There are no rules, AFAIK, about how to build perceptual intent. The idea is to tone down some colors that are in principle within gamut, in order to differentiate them better from colors that were originally out of gamut before being brought into CMYK. But there is no set definition of how far to go (how many in-gamut colors should be toned down) and no definition of "tone down." It's all up to the author of the profile.

3. Why do these 2 profiles appear to handle a Perceptual change so
differently?

Assume for the sake of argument a bright blue. The most brilliant part of the object will come into CMYK with essentially 0% yellow, 0% black, and 100% cyan, with magenta up for discussion. Relative Colorimetric intent will make a lot more than the most brilliant part of the object come in at these numbers, so there will be a loss of detail as previously differentiated colors assume the same values.

Perceptual intent says that we should try to differentiate these less brilliant blues, as well as those bright blues that otherwise would be in gamut. There are three ways of doing this: 1) contaminating these blues by adding black and/or yellow; 2) making the blues lighter by reducing cyan; 3) some combination. Options 1 and 3 are generally preferable. They result in a darker image. There is, however, no law against a vendor adopting option 2.

4. Does this difference suggest any changes in workflow when
converting images?

Other than being aware of the characteristics of the individual profile, no.

Looking only at the preview, it makes the most sense to me to bring
as many colors as I can into gamut and *always* use Relative. If
important colors die when I try to bring them into gamut I'll leave
them alone and try to improve the interpretation myself in CMYK. Is
this a reasonable workflow? Are there times when it's better to use
Perceptual?

I think that you have the right approach. I don't have an objection to using Perceptual but have never found it to be worth the bother. Usually if I want to tone down in-gamut colors I want to do so in a BIG WAY. That would mean doing the work to bring everything within gamut in RGB or else committing to constructing channels in CMYK, just as you suggest. None of the perceptual profiles I've seen are dramatic enough to accomplish these things, probably because they would make more typical pictures look ridiculous.

Dan Margulis