[argyllcms] Re: Rendering intent?
- From: Florian Höch <lists+argyllcms@xxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Mon, 25 Nov 2019 18:39:48 +0100
Am 25.11.2019 um 17:49 schrieb Yves Gauvreau:
I don't like the idea of just decreasing the saturation at least from a
numerical perspective. From what I've read we are not that good at
discerning colors, it takes serious effort on our part and practically
laboratory conditions to discern colors with less than 2-5 DE??
1 dE is considered to be close to a JND (just-noticeable-difference),
but it also depends on the dE formula used (de 2000 will have better
performance and relation to how we perceive color than, say, dE 76 due
to how the L*a*b* colorspace behaves, which has some weaknesses when it
comes to perceptual uniformity, especially in certain hue regions like
blue/purple).
If this is right, just changing saturation would probably
not result in optimal result especially at low L* for example.
The order of importance (most to least important) I would assume for a
"colorimetric" rendering is
Lightness > Hue > Chroma ("Saturation")
or maybe
Hue > Lightness > Chroma ("Saturation")
There's been research on this (lightness and hue are more important than
saturation, e.g. Pariser/Hoshino/Berns), so it seems well-established.
You only have (limited) control over this with cLUT profiles.
With matrix profiles, you're stuck with whichever clipping method the
CMM uses (probably simple clamping of extrapolated values that are out
of gamut, so this can lead to extreme hue shifts).
This should be especially true if the shortest path
also provide a less discernible color difference, doesn't it?
See above.
Florian.
Other related posts: