Gerhard Fuernkranz wrote:
I've seen profiles with S-shaped input tables for the a* and b* channels of the B2A luts (I don't remember exactly, but I think these were profiles made by ProfileMaker). This results in higher resolution for low-chromatic colors, and a lower resolution for highly chromatic colors (many/most of them are usually out-of-gamut anyway). Argyll profiles are apparently not built in this way.
The facility is there (#undef NO_B2A_PCS_CURVES in profile.c), but in my experience, this seemed to make B2A accuracy worse, hence it being turned off by default. I haven't determined why. The A2B table by default uses PCS curves, but it is rare that they do much. I think that for extreme dot gain, the per channel device curves are of more relevance, and Argyll will generate these for both the A2B and B2A tables by default. You can use the -ni flag to profile to turn this off if you wanted to experiment with how much effect these curves are having.
A brief look at the profout.c reveals that -qu implies 23 A2B CLUT grid points (for CMYK), and 45 B2A grid points. But I'm not sure whether you'll have enough patience for -qu :-)
You need a generous amount of memory too.
Graeme Gill.
