[argyllcms] Re: XYZ/LAB to RGB

  • From: Klaus Karcher <lists@xxxxxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 08 Jan 2010 21:18:58 +0100

Pascal de Bruijn schrieb:

Well I'd like to try to produce something like a ColorChecker at home.
Getting as near to the real thing as possible.

So I'd like to add the ColorChecker patches during profiling to make
the profile more accurate on those patches.

So if I add their RGB values to the profiling patches, and then use
the same RGB values to print the target with the profile, I should get
in the ballpark of a real ColorChecker right?

Or am I missing something?

Well you should get an approximate /metameric/ match to the CC -- but no /spectral/ match. I.e. your CCC (ColorCheckerClone ;-) will match the original largely as long as /you/ (a "Standard Observer") compare them under the illuminant the metameric match has been calculated for (typically D50) -- but they don't match for any other illuminant or observer, i.e. your camera (no "Standard Observer") will "see" differences between CC and CCC even when you see none.

A little mind game: When you produce 5 "perfect" CCCs with 5 different printers (ink sets) and use them to profile the same camera, you'll get 5 different results (even when you see no differences between your CCCs). The difference between these results is a measure for the quality of your camera: the smaller the variations, the better your camera fulfills the Luther condition. The Luther condition states that for perfect color fidelity a color camera's spectral sensitivities must be linear combinations of those for the Standard Observer.


In the ti1 files targen generates there must be some correlation
between the RGB and XYZ values?

You could use a preliminary profile and xicclu to calculate the approximate RGB values for the CC in your printer's RGB space and add them to the next test chart to refine the profile in this regions. You could also add some additional points close-by your target points to boost their weighting and to average out printer and measurement variations -- but at the end of the day it's rather irrelevant how accurately your CCC matches the original -- you have to measure the actually obtained values of your CCC anyway and the accuracy of this measurement determines the quality of the target (as you use the actually printed and measured values to build the camera profile).


... and keep in mind that your target will always be biased towards the inkset of your printer.

Conclusion: the spectral properties of the target and the accuracy of the corresponding measurements values specify the quality of your target. The (metameric) similarity to the original CC is negligible.

Klaus

Other related posts: