Gerhard Fuernkranz wrote:
I've noticed, that V0.53 seems to choose a too light "black" for perceptual intent (noticeably lighter than the darkest reproducible neutral L*, at the relcol. [a*,b*]==[0,0] locus), which also results eventuelly in a "washed out" or low-contrast appearance. This seems not to happen in previous versions - their perceptual intent renders black darker. I haven't yet attempted to analyze the root cause. Might it be related to the skewed perceptual gray axis issue, I had discussed with you? Did you probably already include some changes regarding this issue in 0.53, which might lead to the behaviour we see now?
I don't think I made any changes in that area, except for the attempt to clip the black point from profiles that don't have a black point tag, to their gamut. Maybe it was that that caused a problem ? (I've discovered that change wasn't effective. V0.54 will fix it properly.)
Another issue in the current code, is that there's no way to specify an ink limit for the source profile of icclink or profile. This may then give it an unrealistically low black point. (I haven't noticed this much up to now, because I have mainly been testing RGB->CMYK for perceptual intent.) I'm right at this moment working on some new code to get the ink limit defaults to be more sensible ....
I'm doing some more extensive testing of different profile combinations, so I hope these problems will have disappeared from V0.54.