Hi Elle,
RGB spaces may or may not have centered neutral axis. It is a good
premise when you are defining a device space for a physical device
(i.e. a printer or a scanner/camera). But the RGB color model does not
require that. RGB color spaces implementing absolute colorimetric
intent does not behave in such way, to give you a common example.
Centered neutral axis is, however, a very desirable attribute and
most if not all exchange RGB color spaces as well as many device
spaces are using this requirement. This is on color spaces, still
nothing to do with ICC profiles.
Then, when we use ICC profiles to describe color spaces, there are
some limitations that the ICC technology and file format do impose:
First and most important is the quantization. The "normal" profiles
(as there is a floating extension largely unused) have a maximum
precision of 16 bits. That means, no matter what you do, you always
have some uncertainty over the third decimal. That quantity is far
below the JND (just noticeable difference). This is not a big issue
since ICC profiles are intended for human vision, they use spaces like
XYZ or Lab and therefore quantities below the human acuity are not
important.
Because that, a perfect match is not guaranteed. Some widespread
profiles have gaps. Take the usual "sRGB Color Space Profile.icm"
which is a sort of "de facto" standard:
transicc -i "sRGB Color Space Profile.icm" -o *Lab
LittleCMS ColorSpace conversion calculator - 4.2 [LittleCMS 2.05]
Enter values, 'q' to quit
R? 255
G? 255
B? 255
L*=99.9988 a*=0.0188 b*=-0.0173
Yes, that is right! There is some chroma on pure white and it is
located slightly below D50. But still, it does work for many people,
in fact, very few people knows about this mismatch and they are
equally happy using this profile.
Now for your questions:
> First, when making a profile, is there any reason to
> use official D65 and D50 white points that produce profiles that
> aren't color-balanced?
The ICC spec defines values for D50, if you want to fully adhere to
the spec, you have to use such value. The residual amount of chroma
you get is so low that can be ignored. For D65, I use the CCT
black-body radiator formula on 6504K as described by CIE. Really,
should be around 6503.6K
> Second, for an 8-bit image, perhaps it doesn't matter whether an ICC
> profile is perfectly color-balanced or not. Does it matter for the
> increasing number of image editors and raw processors that work with
> 32-bit linear gamma floating point images?
This is the whole point of ICC floating point extension, which does
not suffer from this limitation
> Third, why the discrepency between xicclu and transicc values?
Small discrepancies (below second decimal) may be explained by
different rounding strategies. But you are just checking matrix-shape
profiles which are a minimal set of the huge variation of possible
profiles. In general, I have found xicclu and transicc to be highly
consistent. Much more that Adobe and ColorSync, for example.
Regards
Marti Maria.
Quoting Elle Stone <l.elle.st...@gmail.com>:
> My apologies in advance for a long preamble to three short questions!
>
> It seems to me that a fundamental requirement of an RGB working space
> is that it be color-balanced, meaning that if R=G=B, then in CieLAB
> space a=b=0 (the color is completely neutral gray). Also, for an 8-bit
> image, if R=G=B=255, then LAB L* should be 100 on a scale from 0 to
> 100.
>
> The "sRGB.icm" profile that comes with Argyllcms *is* color-balanced.
> Checking "sRGB.icm" using transicc and xicclu, there is a small
> discrepency between transicc and xicclu values:
>
> $ transicc -w -i sRGB-argyll.icc -o *Lab -t1 -c0
> R? G? B? 255 255 255 L*=100.0000 a*=0.0005 b*=-0.0004
>
> $ xicclu -ir -pl -s255 sRGB-argyll.icc
> 255.000000 255.000000 255.000000 [RGB] -> MatrixFwd -> 100.000000
> 0.000000 0.000000 [Lab]
>
> I wrote a program that uses lcms (v2.5) to make 3 different V2 sRGB
> profiles. Whether the resulting profile is color-balanced depends on
> the values used for the D65 white point. The profile that uses the D65
> value from the Argyll sRGB profile is *almost* color-balanced. But the
> profiles that use the D65 value from the lcms API are not
> color-balanced. This is true whether the profiles are V4 or V2, but
> the examples below are V2 so that xicclu can also read the profiles.
>
> cmsCIExyY d65_lcms = {0.3127, 0.3291, 1.0}; /* from
> lcms 2.5 API */
> cmsCIExyY d65_argyll = {0.312700492, 0.329000939, 1.0}; /*
> argyllcms srgb white point */
>
> Code for three sRGB profiles:
> 1. whitepoint = d65_argyll; /* using d65-argyll */
> primaries = srgb_primaries;
> tone_curve[0] = tone_curve[1] = tone_curve[2] = srgb_parametric_curve[0];
> profile = cmsCreateRGBProfile ( &whitepoint, &primaries, tone_curve );
> cmsSetProfileVersion(profile, 2.1);
> filename = "sRGB-V2-d65-argyll.icc";
> cmsSaveProfileToFile(profile, filename);
>
> 2. whitepoint = d65_lcms; /* using d65_lcms */
> primaries = srgb_primaries ;
> tone_curve[0] = tone_curve[1] = tone_curve[2] = srgb_parametric_curve[0];
> profile = cmsCreateRGBProfile ( &whitepoint, &primaries, tone_curve );
> cmsSetProfileVersion(profile, 2.1);
> filename = "sRGB-V2-d65-lcms.icc";
> cmsSaveProfileToFile(profile, filename);
>
> 3. hprofile = cmsCreate_sRGBProfile(); /* cmsCreate_sRGBProfile */
> cmsSetProfileVersion(hprofile, 2.1);
> filename = "lcms-cmsCreate_sRGBProfile.icc";
> cmsSaveProfileToFile(hprofile, filename);
>
> 1 above, using the same D65 value as the Argyllcms "sRGB.icm",
> produces a profile that is almost identical to sRGB.icm and is almost
> color-balanced:
>
> $ transicc -w -i sRGB-V2-d65-argyll.icc -o *Lab -t1 -c0
> R? G? B? 255 255 255 L*=100.0000 a*=0.0005 b*=0.0008
>
> $ xicclu -ir -pl -s255 sRGB-V2-d65-argyll.icc
> 255.000000 255.000000 255.000000 [RGB] -> MatrixFwd -> 100.000000
> 0.000000 0.001233 [Lab]
>
> 2 and 3 above, using the D65 value from the lcms API, produce almost
> identical profiles that are not color-balanced. Checking
> "lcms-cmsCreate_sRGBProfile.icc":
>
> $ transicc -w -i lcms-cmsCreate_sRGBProfile.icc -o *Lab -t1 -c0
> R? G? B? 255 255 255 L*=100.0006 a*=-0.0020 b*=0.0018
>
> $ xicclu -ir -pl -s255 lcms-cmsCreate_sRGBProfile.icc
> 255.000000 255.000000 255.000000 [RGB] -> MatrixFwd -> 100.000590
> -0.002543 0.002250 [Lab]
>
> Similar results obtain with other D65 profiles (eg adobe_compatible)
> and also with D50 profiles like widegamut and prophoto, where again
> the difference between color-balanced and not color-balanced depends
> on using a white point that isn't exactly equal to official D50 white
> point values.
>
> Three questions: First, when making a profile, is there any reason to
> use official D65 and D50 white points that produce profiles that
> aren't color-balanced?
>
> Second, for an 8-bit image, perhaps it doesn't matter whether an ICC
> profile is perfectly color-balanced or not. Does it matter for the
> increasing number of image editors and raw processors that work with
> 32-bit linear gamma floating point images?
>
> Third, why the discrepency between xicclu and transicc values?
>
> Thanks in advance for insight on this issue of color-balanced profiles!
>
> --
> Elle Stone
> http://ninedegreesbelow.com - articles on color management & open
> source photography
>
> ------------------------------------------------------------------------------
> Learn the latest--Visual Studio 2012, SharePoint 2013, SQL 2012, more!
> Discover the easy way to master current and previous Microsoft technologies
> and advance your career. Get an incredible 1,500+ hours of step-by-step
> tutorial videos with LearnDevNow. Subscribe today and save!
> http://pubads.g.doubleclick.net/gampad/clk?id=58040911&iu=/4140/ostg.clktrk
> _______________________________________________
> Lcms-user mailing list
> Lcms-user@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/lcms-user
------------------------------------------------------------------------------
Learn the latest--Visual Studio 2012, SharePoint 2013, SQL 2012, more!
Discover the easy way to master current and previous Microsoft technologies
and advance your career. Get an incredible 1,500+ hours of step-by-step
tutorial videos with LearnDevNow. Subscribe today and save!
http://pubads.g.doubleclick.net/gampad/clk?id=58040911&iu=/4140/ostg.clktrk
_______________________________________________
Lcms-user mailing list
Lcms-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/lcms-user
