[argyllcms] Re: Perceptual intent

  • From: Florian Höch <lists+argyllcms@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 18 Dec 2014 17:53:20 +0100


Am 18.12.2014 um 16:17 schrieb robert@xxxxxxxxxxxxxxxxxx:
> Input Lab:
> 0 0 0
> 5 0 0
> 10 0 0
> 15 0 0
> 20 0 0
> 25 0 0

An L* of 0 will most certainly clip when fed through the colorimetric
B2A table of any output device (you can add the -a switch to the xicclu
commandline to see when values are clipped).

> There is a huge difference in the darks between the two and I don't
> understand why the perceptual darks are so much lighter that the relative
> darks.

Unlike the colorimetric B2A table which usually clips if BPC is not used
(note that xicclu has no notion of BPC), the perceptual B2A table maps
the lightness axis, i.e. an input L* of 5 will not result in an output
L* of 5, and so the two xicclu results are not really comparable. If you
want to compare the xicclu colorimetric to perceptual results, you'd
need to at least offset & scale the input L* values by the black point
for (only!) the colorimetric xicclu lookup. E.g:

Input Lab:
3.06 0 0
8.04 0 0
12.36 0 0
16.70 0 0
21.27 0 0
25.98 0 0

(I obtained those values by doing simple offset & scaling in XYZ, aka
BPC, based on the black point L*a*b* of 3.06 -0.12 1.19 from an earlier

Note though that I'm not sure how useful such a comparison is in the
first place, as the very purpose of the perceptual B2A table is to do,
well, a perceptual mapping, i.e. aimed at being visually pleasing.
Constructing such a mapping (i.e. the job of the profiling software) is
both science and art, and that's why results can vary so much between
different profiling softwares.

So if you do a relative colorimetric and perceptual rendering and the
two are different (quite likely), then the only thing that this really
tells you is that they are different ;)

> Sorry, it's here:
> http://www.irelandupclose.com/customer/argyll/perctest.jpg

The (presumably?) disabled black ink simulation in Photoshop makes it
hard to visually compare the mapping near black.

I'd suggest enabling "simulate black ink" in Photoshop's softproof
settings. What you'll likely observe is that the perceptual rendering
will preserve more detail near black than the relative + BPC one (and
definitely if you disable BPC for the relative rendering, because then
it'll just clip).


Other related posts: