H. Teeuwen wrote:
I have just started experimenting with Argyll cms and I'm know redoing some profiles that I initially created with Profilemaker. As Argyll cms allows to specify an icc profile to apply gamut mapping to perceptual and saturation B2A table, I was wondering how to go about when creating an weighted averaged matrix profile myself.
There's no particular facility in to do this currently. I'm not quite sure off hand even how to go about it for a general gamut as it is handled in Argyll since it uses a triangular gamut surface representation. One approach would be to first deal with any differences in white point, and then for each triangle vertex locate the equivalent points on the other gamuts, average them, and use the averaged surface points to construct a gamut surface. I suspect from the nature of the colorspaces mentioned, that "averaging" has been done simply by averaging the colorant locations to construct a single matrix style profile.
- Supposed that I'm able to create/tweak such profiles myself, could I just average the white point and primaries of two working spaces (as specified in profile which is always relative to D50) and include and sRGB or L* tone response curve? - How do I average white point and primaries?
There's only one logical answer - pick a colorspace and average the coordinates. Of course you could apply other 3D geometry in the colorspace if you liked as an alternative (ie. center of the smallest volume enclosing the points etc.) - the trick is to come up with some rationale for your choices.
- For the weight, could I use something like the relative occurrence of each of the working spaces? For example, suppose that 75% of images that will be used with a specific profile is in sRGB, and 25% is in aRGB, and then use these percentages for the weight?
Sure. You just need a rationale for the weights. Graeme Gill