[argyllcms] Re: inconsistencies in what different programs show

  • From: Michel Joly de Lotbinière <michel@xxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 18 Mar 2010 16:48:03 -0400

There are people on this list much better qualified to answer your
questions, but I will take a stab at them (I'm talking only about
ArgyllCMS produced profiles):

1) Differences in color management engines used by different
programs/operating systems: This is a real issue.

I use Windows 7 (but the same applies to earlier versions, I recently
took advantage of Lenovo's reduced price OEM upgrade offer from Vista,
now expired,  for those purchasing a machine since June of last year)
and a photo editor (Picture Window Pro, dl-c.com) that allows
switching between the built-in Windows ICC color management and

With matrix profiles (whether 1xcurve or 3xCurve) there is a subtle
difference: shadows are just a tiny bit darker with Windows CMM.

With LabLUT profiles, there is a difference too, but with little
experience I haven't noted exactly what it is. Generally LabLUT
profiles lighten the highlights comparatively to to same data used to
produce a curve+matrix profile.

With XYZLUT profiles, things get weird: on progressive step wedges
(such as those in test images downloaded from lagom.nl). LittleCMS
produces irregular gradations, while Windows CMM displays quite
regular gradations.

There are screen captures at
but don't attach importance to the exact colors or gradation since I
had a slightly different display LUT loaded when I grabbed these--but
the same thing is observed with the actual XYZLUT profile loaded in
the display.

I thought at first this might be an artifact of too low a number of
patches used to produce the measurement file, but profiles produced
using Florian Hoech's "massive testchart" (2000+ patches) exhibited
the same behaviour. I don't know what to make of it.

2) Firefox 3.0.X used the LittleCMS engine, but color management was
not activated by default. Once activated and an appropriate actual
display profile specified (sRGB was the fallback), all embedded
profiles were honored and all other colors were mapped from sRGB space
using perceptual intent (i.e. colorimetric for matrix profiles and LUT
profiles without a perceptual table). This set of defaults actually
meets the W3C WWW color management standard, apparently, and thus
Firefox 3.0.x was the only "color correct" browser (Apple's Safari is
a close second, but they have chosen to not transform untagged
graphics--they are assumed to be already in monitor color
space--because plugins such as Adobe Flash don't do color management
and there would be visible inconsistencies).

With Firefox 3.5.x and 3.6.x, Mozilla has developed its own CMM,
"Quick CMS". It supports only version 2 matrix profiles, it is
milleseconds faster for each image, and it is code that Mozilla
completely controls. But you really have to wonder if jettisoning the
whole color management knowledge + experience embodied in LittleCMS
was a good idea:  you can get an idea of the current bug list at

But they have also added the capability of specifying the CMM
rendering intent and whether or not it applies to all or just tagged
graphics. One step forward, two back, etc. See
for all the knobs to twiddle.

I'm remaining with Firefox 3.0.x until the situation is more sorted
out. And anyway, the quality of my notebook's LCD is such that
ArgyllCMS's 1xCurve+MTX profiles do a very good job in creating a more
tractable display quality.

3) Dealing with uncalibrated screens: I simply load a linear LUT in my
display and use an umanaged application to view the photos, just to
make sure that none of my edits depend on having a calibrated display
(deep shadows and highlights are areas to watch). I suppose someone
who is sending images for printing/publication purposes has to find
out what will happen at the other end, to determine if everything
should be converted to sRGB before sending. At least with physical
printing you have a hope of consistency in results--there is no such
hope online.

Anyway, these are my answers, such as they are...

On Wed, Mar 17, 2010 at 12:32 PM, Leonard Evens
<len@xxxxxxxxxxxxxxxxxxxxx> wrote:
> I have  two computers presently, each dual booting Fedora 11 and some
> version of windows.  I've used argyllcms to calibrate/profile my
> monitors under Fedora and the Xrite software to do so under Windows.  (I
> haven't worried a much about getting it right under Windows.)

Other related posts: