Alastair M. Robinson wrote:
The one scenario where you might want to do this is if you're using tiffgamut to compute a given image's gamut with a view to making a devicelink specifically for that image, and that image happens to be in CIELab space.That, I think, is the scenario Nikolay had in mind.
Ok, yes you're right. I thought it might just work, but there are several problems currently. The problems hint at some conceptual issues as well. Problems that I notice are: The image gamut has no white and black point. I can simply assume in tiffgamut that an Lab image has white and black points of 100,0,0 and 0,0,0, but there is a trap waiting if someone converts from a device space straight to Lab and doesn't ensure that the conversion also expands and maps the device space white and black to 100,0,0 and 0,0,0 There is a bug in tiffgamut that means that selecting CIECAM02 output for compatibility with collink is being ignored. The resulting gamut will be in L*a*b* space. The gamut mapping code takes the number of guide points from the source gamut surface, and this depends on the surface complexity (number of facets) and shape. For the Lab space source, the surface is a cube with only 8 vertices so the number of guide points, or for Jab it is stupidly large with facets in places that aren't relevant to the image gamut. It should be possible to fix these various problems for the next release. Graeme Gill.