[argyllcms] Re: ArgyllCMS: shadow details and blue color issues. What I'm doing incorrectly?

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 10 Oct 2012 23:40:48 +1100


> The results were not acceptable when I use the perceptual intent due to a 
> very muted or
> washed blue color and a slight magenta cast. Blacks are deeper blacks.

It's always hard to investigate such issues without all the details
of the workflow and being able to examine the resulting print, but
at least in the softproof simulations of the profile, I'm not sure I
can really reproduce this. While the perceptual profile (as expected) has 
less saturated blues (see <http://www.argyllcms.com/RGBsurfall_pv.tif> than the
colorimetric intent (see <http://www.argyllcms.com/RGBsurfall_rv.tif>,
they are not what I'd call washed out, and the neutrals are not obviously 

Note though that it depends to some degree what you are looking for.
The nature of a print device is that there is a severe trade-off between
saturation and lightness in blue. The Perceptual intent (by the tuning I've 
is lighter and less saturated than the clipping of the colorimetric intent,
and (to my taste) is visually closer to the very intense and light
color of the source AdobeRGB blue (Helmholtz-Kohlraush effect). The colorimetric
reproduction is certainly more saturated, but is overly dark in my opinion.

> If I print using the same profile, and the same paper, but using the relative
> colorimetric intent, the blue color is perfect (not washed or muted), but the 
> blacks
> are more dark greys than blacks. Also I loose the shadow detail.

That's a consequence of the device having a very non-neutral black
point, and it being a colorimetric intent. "Colorimetric" strictly
implies point by point absolute minimum color error, and the shortest
distance from the target of perfect black to the gamut surface is dark
blue/grey. A darker mapping would have a higher saturation error, and
a higher total delta E.
In contrast, the perceptual mapping makes more effort to map
the neutral axis and luminance ranges. You could try a "la" - Luminance axis
matched Appearance intent for the perceptual table (-t la)as a compromise
between the two. Another option is to use one of the two saturation intents, 
should give good blacks and more saturated (and possibly darker) blues.

> Thinking that I will improve blacks and shadow details on relative 
> colorimetric, I have
> do charts with more patches: 396 patches with 48 greys and 3 whites, 1188 
> patches with
> 128 greys and 3 whites and 1980 patches with 256 greys and 3 whites.

Since the black is a result of the device behaviour near black and the intent,
increasing the number of patches is not likely to change anything.

> *colprof -v -qu -iD50 -o 1931_2 -S AdobeRGB1998.icc -cmt -dpp -D"Description 
> Profile"
> -O"NameProfile.icm" Argyllv140_396parxes48Gris3Blanc

I'm quite seriously thinking of changing the documentation for the "-q" flag:

-q fmsu         Speed - Fast, Medium (def), Slow, Ultra Slow

to try and dissuade people from "turning it up to 11" without
good reason. (<http://en.wikipedia.org/wiki/Up_to_eleven>).
Your device seems reasonably well behaved, and I would think
needs no more than -qm, or -qh at the most.


Graeme Gill.

Other related posts: