[argyllcms] Re: Monitor calibration: blue turns purple

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 26 Nov 2012 06:59:53 +1100

Federico Cozzi wrote:
> I own a laptop (Acer Aspire 4810TG) and a Huey Pro colorimeter. I
> calibrated my monitor with the included utility only to find the
> infamous "blue turns purple" problem:
> http://www.brucelindbloom.com/index.html?MunsellCalcHelp.html

        there are many "blue turns purple" problems. One
is the above gamut mapping in a non-hue uniform space. Another
is using the wrong (or no) chromatic adaptation matrix. Others
are simply profile inaccuracies due to instrument limitation
(ie. colorimeter spectral inveracities or model fitting
residuals). So what leads you think that gamut mapping in a
non-hue uniform space is the source of the particular effect you see ?

> Since in the past I used argyll CMS to calibrate a slide scanner
> (Reflecta Digitdia 5000) and got wonderful results compared to
> Silverfast AI, I was wondering whether Argyll CMS could help me in
> this case as well.


> I was intrigued by Uniform Perceptual color spaces e.g.
> http://www.brucelindbloom.com/index.html?UPLab.html.
> In my limited understanding of color management, I thought that if I
> could profile my monitor using a uniform perceptual color space as
> "profile connection space" instead of the standard CIE lab, the "blue
> turns purple" problem could be mitigated.
> Does this make any sense?

Not really. The PCS (which is mandated by the ICC spec. to be either XYZ or
L*a*b*) is used for storing colors in the profile. Profile
creation and/or gamut mapping can be in any colorspace you
like that is derivable from the PCS. Some CMM's may use
the PCS for gamut mapping or clipping, but they are not
forced to do so.

> Is something like that achievable with Argyll CMS?

Argyll already uses CIECAM02 by default for gamut mapping,
which is known to be more hue uniform than L*a*b* space.
But to take advantage of that, you need to be using
the gamut mapping. That means using cLUT based profiles,
specifying the source colorspace you will be using
the profile with at profile creation time (unless you
use collink of course), and invoking the appropriate
intent (ie. perceptual). Not all CMM's cope with this,
and cLUT profiles can have some subtle disadvantages at times.

> What else could I try, apart from buying a better monitor?

As always, some careful detective work is needed to track
down the source of such problems. Is it the instrument ?
Is it the profile ? Is it the workflow ? Is it some element
in the workflow ?

Graeme Gill.

Other related posts: