Graeme, thank You for paying your attention! Graeme Gill wrote: > Nikolay Pokhilchenko wrote: > > Hmm... Which way the ArgyllCMS determinate the printer black point for an > > RGB printer? > > There's no such thing as an RGB printer, so in fact there is already a > transform that > someone has set up to convert from RGB to the actual printer colorant. So we > assume that whoever set this up knows what they were doing, and RGB 0,0,0 > must be the black point. I've noticed, that for RGB profile with not-neutral RGB 0,0,0 point, the result still is OK, because the tone change from white to black is stretched. > > In my practice I can deal with different quality CMYK inks, so the "K" inks > > hue may > > be quite high. > > You mean it's not very neutral ? Yes, It may be rather non-neutral, but still dark. > > I think it will be more robust to determine the black point only by gamut > > envelope (boundaries) and to pay respect to vision adaptation to the > > printer black. > > I don't know of any way of determining black reliably from the gamut > boundary. The lowest > L* could be very non-neutral. > > There is no such thing as black adaptation. ... > Certainly one of the alternative is to look for the darkest black that has > the same > hue as the paper. This should make the white to black axis look uniform. As I mentioned above, the minor tone shift may be acceptable. In some cases even the heavy tone shift at blacks can be preferred in an effort to the contrast. > > Another variant is to determine the point at the straight line between > > Jab 0,0,0 and 100,0,0 with J=Jmin, where Jmin - the minimum achievable J > > value for given > > gamut with given TIL (without > > regard for hue). The optimal black point will be the closest for the point > > determined. > > Hmm. That would make an interesting not-quite-neutral "bend" target. Yes. I think there should be a compromise between the dark neutrality and the contrast. I think the two values: the "lightness" and "unwanted tone shift" should be computed in perceptual space for every candidate point, then the best point by lowest sum of this values should be selected.