On 12 Nov 2009, at 16:03, Graeme Gill wrote:
Idea Digital Imaging wrote:
I tried adjusting the knee setting to 0.5 but it made big changes to the lightness of saturated colours and very little difference to the skin tones. I assume that the "knee" is a point where the gamut compression stops/starts being linear and that changing it in isolation isn't likely to achieve all I want?That's not what I would expect. The gamknf controls how much the gamut surface compression/expansion penetrates the volume. A value of 0 gives almost uniform compression/expansion from the surface to the center of the volume. A value of 1.0 will almost completely preserve the source gamut that falls within the destination volume, and clip colors outside that to the destination surface.
I think the reason it made little difference to the skin is because the skin was all in gamut in the first place and the problem seems to be restricted to a relatively small area in my tests.
What is getting lost is the red "blush" in the light skin tones (on the petals it was a hint of a warm glow). If I overlay a strip of the separated image over the same file that has been converted with relative intent and black point compensation in Photoshop (still there using perceptual too) you can see that this delicate blush is going neutral:
Perhaps it's an issue with the profile itself -- ISO Coated v2 (300) -- or with the viewing conditions etc which only appears when you try gamut mapping? There are a few too many variables in this process.
I don't think I can address your particular issues until I get back home and can evaluate things under more controlled conditions.
My needs and feature requests are niche so can appreciate that you've got lots more pressing/interesting stuff to do.
Regards -- Martin Orpen Idea Digital Imaging Ltd