Martin Huber hat geschrieben:bkh hat geschrieben:let me add that, if PL's blending modes are supposed to be similar to those in PhotoShop (the manual doesn't explicitly say so), then your example exhibits a bug in PL.
Why do you think so? I tried that example with (an old version 6) Photoshop, and the result was the same as in PhotoLine.
bkh hat geschrieben:In fact, since saturation and lightness in your example are constant,
The way lightness is defined for the blend modes, lightness is not constant. Therefore the blended result is partially darker.
Martin
I see, luminance/lightness in that blend mode is not supposed to be the same as lightness as the "L" in HSL (since H and S seem to come from that definition, I expected L to belong to the same color model), and neither the L from CIE Lab, or the intensity from HIS (= HSL?) which Curves... uses.
I don't have a recent version of Photoshop available to try, but at least
http://www.nathanm.com/photoshop-blending-math/ seems to indicate that it uses the plain HSL color model for blending. This may, of course, be wrong, but to me, this seems natural and sensible.
In any case, the luminance used in the blending has an enormous drop at pure blue, and a smaller drop at pure green. See the attached example file, which shows the luminance as used by the blending mode along with the L channel from CIE Lab, which PL apparently uses in its grey mixer (the L in HSL would just be 50 % grey). These drops (which do not reflect visual changes in luminance) seem to cause the rapid color changes in Loadus' example, and my guess would be that these are visible (and annoying) in many other instances as well. So, in my opinion, the luminance function used in blending mode is broken. Why not replace it by L in HSL (i.e., the average of max(R,G,B) and min(R,G,B)) or L as in CIE Lab (the latter might be too slow...).
Cheers
Burkhard.