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

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 06 Aug 2012 13:58:05 +1000

Alexey Blinov wrote:
> Graeme, I'd like to summarize "for dummie" out here.
> Since sRGB assumes the display at D65, and V2 profiles from color.org use 
> D50, they are not
> supposed to be used as input sRGB profiles. That is, the kind of profiles 
> that specify meaning
> ("units") of an RGB number triple in an image file.

Hmm. I guess I wouldn't put it that way. If all you are using is relative
colorimetric intents, then there is no difference between profiles that
have D50 as their wtpt and those that have their actual white point as the wtpt.

If you use absolute colorimetric, then there is a difference:

CMM intent      D50 wtpt + chad         Real wtpt, no chad

Relative        Relative                Relative
Absolute        Relative                Absolute
Special Abs.    Absolute                N/A

The "Special Abs." intent represents a special CMM absolute mode
that applies the reverse of the chad tag to recover the true display
white point. I don't know of a CMM that implements this.

> Instead, they are provided partly for a specific soft-proofing scenario with 
> a dumb CMM.

That was just one explanation I was given, that doesn't make it the actual
reason that the ICC have gone this direction (remember that the ICC is a 
of people representing companies, and its composition has changed over time).

One other technical reason that they may have gone in that direction is that
the ICC spec. uses a "wrong Von Kries" chromatic transform for adapting between
the white point and D50. For most print profiles this error is not significant,
but for a display profile it is significant. The Microsoft/HP sRGB and AdobeRGB
used a Bradford chromatic transform for this, which is not strictly ICC, but
is the correct technical approach. The ICC, rather than addressing this properly
by allowing something other than "wrong Von Kries" to be used for this, instead 
the chad tag. So to be strictly ICC compliant and still use a Bradford 
you get forced into using a D50 wtpt tag and disabling absolute colorimetric.
[I think that sRGB and AdobeRGB preceded the introduction of the chad tag.]

Note that what the ICC should have done (and could still do!) is introduce a tag
that looks a lot like the chad tag that allowed the profile to specify what
chromatic adaptation transform should be used for its absolute <-> relative
white point transform, avoiding the "wrong Von Kries". The chad tag would then
just be used for the rare case of a print profile measured under a non D50

> For input sRGB profile, one must use a profile from some reliable source (like
> argyll/ref/sRGB.icm). 

I don't see any down side to using an original Microsoft/HP sRGB profile or
the Argyll equivalent, since it enables CMM's to use absolute colorimetric,
which is potentially useful for soft proofing, whereas an ICC sRGB profile
removes that possibility with most of the CMM's out there.

> But incorporating input sRGB profile in an image file is pointless because
> all the consumer software already assumes such correct sRGB profile.

I don't know that that is true. The availability of such D50 wtpt profiles from
the ICC and the demise of the Microsoft/HP sRGB website makes that less likely
to be true as time goes on.

Graeme Gill.

Other related posts: