I can't believe that I seem to have gotten the last word on this one.
Come on, Crawford, show me why I'm wrong.
LAB's lightness channel is lighter than a corresponding greyscale composite, yet keeps the same black endpoint. So it stretches out the dark range at the expense of the lighter tones, which get crushed. That's where all the relevant facial detail usually can be found. Your colors are okay; a little yellow for my taste, but I think that's a function of the face being flat overall. I took your image back into LAB and played with the lightness curve and it's impossible to get the subtle shifts that are needed. Posterization creeps in somewhere in the range no matter how slight a move. I forget which chapter it's in, but Dan's book makes that very point: that LAB is better for working contrasts in shadow detail and RGB or CMYK better for lighter detail.
Also, sometimes you want your highlights to move more towards neutral. Lab keeps them colored with imaginary colors, neither a good thing nor a bad thing, just something to factor in. I think with this face, it creates an unreal cast when you try to bring out the highlights in lab.
I don't doubt that it's possible to get a decent result in LAB, but i don't think that type of correction is it's strong point. Heck, even with the high quality images that were posted for your Chapt. 16 summary, after all the good stuff in LAB, it was often worth the trouble to make a stop back in RGB for ae color tweak. I think, given the kind of bi-directional color moves going on here, precision is more easily attained in RGB where the curves do precisely what is called for: lowering one channel in one range, raising another channel in another range.
There are two ways to slide through life: to believe everything or to doubt everything; both save us from thinking.
—Korzybski
I often make subtle RGB tweaks for portraits before going to LAB. I wanted to do that in this case. But something a lot more than a tweak was required here, which is what drove the decision to repair the cast in LAB.
Comments
Come on, Crawford, show me why I'm wrong.
LAB's lightness channel is lighter than a corresponding greyscale composite, yet keeps the same black endpoint. So it stretches out the dark range at the expense of the lighter tones, which get crushed. That's where all the relevant facial detail usually can be found. Your colors are okay; a little yellow for my taste, but I think that's a function of the face being flat overall. I took your image back into LAB and played with the lightness curve and it's impossible to get the subtle shifts that are needed. Posterization creeps in somewhere in the range no matter how slight a move. I forget which chapter it's in, but Dan's book makes that very point: that LAB is better for working contrasts in shadow detail and RGB or CMYK better for lighter detail.
Also, sometimes you want your highlights to move more towards neutral. Lab keeps them colored with imaginary colors, neither a good thing nor a bad thing, just something to factor in. I think with this face, it creates an unreal cast when you try to bring out the highlights in lab.
I don't doubt that it's possible to get a decent result in LAB, but i don't think that type of correction is it's strong point. Heck, even with the high quality images that were posted for your Chapt. 16 summary, after all the good stuff in LAB, it was often worth the trouble to make a stop back in RGB for ae color tweak. I think, given the kind of bi-directional color moves going on here, precision is more easily attained in RGB where the curves do precisely what is called for: lowering one channel in one range, raising another channel in another range.
—Korzybski