Gerhard Fuernkranz wrote:
I'm wondering whether clipping to the nearest point on the gamut surface is optimal, when we are not dealing with a strictly convex gamut surface.
Hello Gerhard, that thought has occurred to me, and I have looked into various approaches of mitigating the issues caused by non-convex surfaces. Within the gamut mapping code itself, there is weighting given to the smoothness of the mapping, and this tends to override any potential discontinuities. Within the colorimetric code, it does strive for a strictly nearest mapping, since colorimetric accuracy is (assumed) to be the priority over smoothness for that particular intent. I did come up with some other ideas too, but haven't bothered trying to implement them, without some example of this being a real world problem. cheers, Graeme.