[argyllcms] Re: ProPhoto to sRGB device link problem

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 25 Sep 2009 08:51:31 +1000

Idea Digital Imaging wrote:
I've made a number of ProPhoto to sRGB device links with collink using a range of different settings.

All of them create images with odd dark blobs appearing in saturated blues.

Yes, I've seen this. The problem is that the ProPhoto primaries
are so way out there, it runs into all sorts of problems.

One is that none of its primaries can be represented in ICC Lab PCS.
They all have a* or b* that exceed the -128 to 128 range:

Red:   0.797668, 0.288040, 0.000000    [Lab 60.608273, 139.168888, 104.497022]
Green: 0.135193, 0.711884, 0.000000    [Lab 87.576439, -186.694184, 150.993860]
Blue:  0.031342, 0.000092, 0.824905    [Lab 0.082699, 90.249204, -172.271644]

In fact the green and the blue primaries are outside the
spectrum locus, meaning that they are imaginary.

Notice that the blue has almost no luminance level!

Given that key technologies such as CIECAM02 color appearance
space were never designed to work on non-real world colors,
attempting to transform the whole of the ProPhoto gamut
runs into issues, particularly near blue.

While I have attempted to mitigate this issue in using CIECAM02
in the next release, the attempt is not entirely successful for ProPhoto.

There are two possible approaches that may work. One is
not to use CIECAM02, ie. try the "rl" intent for instance.

The other (if you haven't been doing so) is to setup the link
with an image specific source gamut as well as the ProPhoto
source colorspace, so that the transformation is not attempting
to map the whole ProPhoto gamut, but just a real world
portion of it. You could even try some other RGB colorspace
gamut, such as AdobeRGB or sRGB as the "image" source gamut.

See if either of those ideas help.
Graeme Gill.

Other related posts: