V2 and V4 profiles - Out of Gamut indicator and review of ICC Profile Creator

[Desmis]

I do not repeat the terms of the same topic in the forum :)

I bring more technical complements.

I will especially treat 2 points mentioned in the forum that seem to me to be the responsibility of the "ICC profiles creator": a) I think it is not a good idea to use RTv4 profile for monitor, so in "Preferences" I disabled this choice

b) differences between AdobeRGB1998 and RTv2_Medium when using as "Monitor profile" If you load in Preferences / Color Management / Default color Profile, and you choose an image with large gamut (flowers...) and you click on "Out of gamut", you can see a difference, big area OOG are the same, but very small dots are present with RTv2_Medium

I have search why ? First I suspect the small differences of rXYZ, gXYZ, bXYZ between the 2 profiles, these differences (very small) are due to differences in media white point with calculations made by LCMS, after correction always same differences in OOG screen After, I suspect the action of TRC, AdobeRGB1998 has a gamma=2.2 (in fact 2.19921875) and RTv2_Medium gamma=2.4 slope=12.92

• I replace gamma=2.4 and slope 12.92 by gamma=2.2 and slope=0, after correction always same differences in OOG screen

  • I suspect, the imprecision of the LUT (4096) and replace LUT by "calculate gamma" with "cmsBuildGamma(NULL, 2.19921875)", and now all works fine.

  In summary, now, AdobeRGB1998 and RTv2_Medium are strictly identical, except of course description, copyright, etc.

  This observation led me to make a complete review of "ICC profile creator", for the most part it concerns ICC v2

  • adjusting media white point values, instead of calculation by LCMS (very very small differences)
  • manual input of rXYZ, gXYZ, bXYZ values, instead of calculation (very very small differences in some cases)
  • add some tags forgotten for "displayclass" whithout effect!!
  • modified Options.cc to take into account RTv2 as reference to copy/generate ICCv2 instead of RTv4, this lead to suppress tag "chad" which is useless for v2, but which was not used and have no effect for v2.
  • all ICCv2 and ICCv4 has been "rewrite" (except RTv2_Beta, I used as reference) to take into account all modifications (calculations,white point, etc.)

  All these changes does not affect usage in "Output Profile" (differences are very very very small)

  All the ICCv2 - except RTv2_Medium - and ICCv4 has by default a TRC gamma sRGB gamma=2.4 slope=12.92 RTv2_Medium as a calculate gamma=2.19921875, in the case of you want to use it for monitor.

  Of course, I have not change others fonctionnality, you can always for ICCv2 adapt the TRC, and for ICCv4 : TRC, illuminant, primaries.

  I push a commit d55d344 with all this changes in branch enhance_icc

  jacques

[Desmis]

I just forgot to change the labels of the "desc" tags, now they are all RTv2_xxxx or RTv4_xxxx

It is suitable to have best quality with simpleprocess.

Now it's done 25eb93a

[Beep6581]

For reference, branch comparison: https://github.com/Beep6581/RawTherapee/compare/enhance_icc

[Beep6581]

I used the branch for a while with no issues.

[Desmis]

@Beep6581 Thank you for testing :)

I just commit a new "normalization" for D50 profiles (Prophoto, Widegamut, BestRGB, BetaRGB). I adjust media white point to Room specifications... very small differences with before :)

And another commit for White point D60, D65 D50 for ICCv4

To note that white point used by LCMS is slighly different that all D50 white point, but very small diffe

I hope there is no bug, and I think I have finished review iccprofilecreator.cc

All these differences are very small, and does not change results (or very tiny !)

jacques

[Desmis]

I have found another mini bug...I will solve soon :)

[Desmis]

Now it is good :)

[Desmis]

if no objections I will merge this branch tomorrow :)

[Thanatomanic]

Just for reference: https://discuss.pixls.us/t/v2-profiles-v4-profiles-and-rts-out-of-gamut-indicator/9734/

No objections for a merge, but mostly out of ignorance 😉

But a question still: @Desmis

a) I think it is not a good idea to use RTv4 profile for monitor, so in "Preferences" I disabled this choice

Are all v4 profiles unsuitable as a monitor profile? If yes, then why exactly?

[Desmis]

@Thanatomanic

For RTv4 and profile for monitor, I don't know why, I find that it does not work properly

@Beep6581 Thank you :)

[Beep6581]

@Desmis since the enahnce-icc branch was merged into dev 152b42125ae3606da08a5d9d0e1e617ceba77d00, shall we close?

[Desmis]

@Beep6581 Please don't close, I think I will make changes in the coming days, but not about 5.5 :)

jacques

[heckflosse]

@Desmis Jacques, my I change the line endings to LF in rtgui\iccprofilecreator.h and rtgui\iccprofilecreator.cc in your branch?

[Desmis]

@heckflosse

Ingo

No problem you can :)

[heckflosse]

@Desmis Jacques, done, also merged dev into your branch

[Desmis]

@heckflosse Ingo, thank you for this work :)

I push in "enhance_icc" a commit with a) some bugs solved ==> when you used "Primaries : Custom" the name of the files did not take account of "custom" b) sometimes Tag "dmdd" which store "gamma and slope" to be reused (iplab2rgb.cc) for ICCv2 did not have good values

And the most change c) Now ICCv2 profiles have the same settings as ICCv4: you can use "Primaries - Custom" and Illuminant as thoses of ICCv4 (Default, D41, D50, D55, D60, D65, D80, stdA)

With LCMS it is impossible , "cmsCreateRGBProfile" only generate ICCv4 I "copy" my spreadsheet, to RT code. Now there is the possibility, if it helps, to generate in RT the RGB-->XYZ conversions, from the primaries (including "custom") and illuminants, for any purpose other than ICC profiles, of course by modifying the code :)

jacques

[Desmis]

I make others modifications :

• for RTv2_Medium, tag "dmdd" was whith wrong values
• if user select TRC = custom and slope=0, now code use a calculated curve instead of LUT (best result)
• This modification is take into account in iplab2rgb if user select RTv2 profiles in output profile

I think now all bugs or improvments are identified and fixed

If no objections, I will merge this branch in dev, tomorrow :)

[Desmis]

@Beep6581 @heckflosse Please can you merge this branch into dev, if of course you agree, because I cannot, my system is broken

I will open an issue to explain

Thank you Jacques

[Beep6581]

@Desmis of course, doing so now.

[Desmis]

@Beep6581 Thank you very much I think now we can close this branch.

And now thanks to @Thanatomanic , my computer works again :)

jacques

[Beep6581]

Okay. Branch enhance_icc deleted, closing issue.

[RawConvert]

@Desmis , hi, I've built Dev tonight and from the above it should have your changes in I think. One point - though V4 profiles have been removed from Preferences, you can still select them via the preview screen. If you do this, then go into Prefs, then click CANCEL, the monitor profile in the preview window is updated!! Yes, Cancel! Good news though is that even with V4, the anomaly of showing the wrong OOG first time through looks to have gone. Thanks for your efforts.

[agriggio]

Hi @Desmis and others:

All the ICCv2 - except RTv2_Medium - and ICCv4 has by default a TRC gamma sRGB gamma=2.4 slope=12.92

wouldn't be better to have linear gamma for the ACES profiles? That's what I would have expected... if not, why not?

[agriggio]

Reopening so that it gets noticed -- I think this should be answered for 5.5

[Beep6581]

I will tag 5.5 tomorrow. Maybe @Desmis is away on holiday. Please shout if you feel action needs to be taken before tagging.

[Desmis]

@Beep6581 I am back at home.... For alberto @agriggio , we can do what we want ...and the principle of this tool is that everybody can do what he want... For RT_v2 and RTv4 of course we can change TRC, but why ? If the answer is yes, in this case we must change RTv2 RTv4 Large with TRC 1.8 RTv2 RTv4 Mdeium with TRC 2.2 and probably others profile (Bruce...???)

But everyone can easily make this change

[agriggio]

@Desmis Jacques, I understand that everybody is free to do what they want, my question was more about what should be the proper default, and IMHO for ACES at least the TRC should be 1.0...

[Desmis]

@agriggio

OK as Prophoto (1.8) AdobeRGB (2.2) But why ? For me no problem to have other TRC.... The best choice for me for all profile (except perhaps for ACES_P0 which has a very big gamut and usages are specific....stock data... in this case TRC have no importance !) would be BT709 (g=2.2 s=4.5) and not sRGB (g=2.4 s=12.92) in summary, if you want to change, any problem for me :)