[argyllcms] Re: wtpt/chad in input V2 sRGB profiles from color.org

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 03 Aug 2012 16:07:44 +1000

Alexey Blinov wrote:
> They say they set wtpt=D50 to eliminate color cast when absolute colorimetric 
> intent is used. To
> me it's vice versa: it introduces a yellow cast, when I set it as (sRGB) 
> profile in GIMP.

Some people at the ICC have a strange point of view with respect to absolute
colorimetric and display profiles. From past discussions it appears that
the move to set display profiles wtpt at D50 (flying in the face of those
millions of sRGB and Adobe RGB profiles out there) was motivated partly
by one sort of soft proofing scenario - soft proofing the difference
between hard copy paper white and illuminant white, but displaying it
relative colorimetrically on a display. Rather than punting this to
the CMM where it belongs (ie., to get a mixed absolute + relative
colorimetric transform, specify it in the CMM that way), they altered
the display profile to make absolute colorimetric == relative colorimetric
so that a dumb CMM would give that result when told to use absolute 

This leaves anyone wanting an actual soft proof out of luck if they are
stuck with the normal selection of intents.

> {r,g,b}XYZ tag values (colorants) from those profiles coincide with those 
> from Argyll's sRGB
> profile. However, iccgamut+viewgam show whitepoints around D50 and D65 
> correspondingly for
> color.org and argyll profiles. So it looks like color.org assumes a transform 
> using chad, making
> white point D65->D50. Why?

It's hard to fathom, and fundamentally wrong. A profile is meant to represent
the characteristics of the device. For print, the illuminant is not part of the 
but can be chosen independently, so it is removed from the profile by using a 
D50 illuminant or chad tag if some other illuminant was used for measurement.

In contrast a display's "illuminant" is fundamentally part of the display,
and therefore the wtpt should reflect the actual media white point.

This madness is now standardised in V4 :-(

Graeme Gill.

Other related posts: