It took me sometime to understand that despite the results on the ColorChecker
SG looked better, some colors outside the chart may be strongly deviated while
creating XYZ cLUT profiles.
Let me resume my findings:
a.. Scanner ICC profile created using “XYZ cLUT” setting on a referenced
ColorChecker SG: Lowest DeltaE on ColorChecker SG patches, strong deviations on
some colors outside the ColorChecker SG with some posterizations and colors
drastically changing hue (some yellow become green and so on). Results may
changes drammatically if for example there is a slight change in the reference
LAB values of some color patches or using a different ColorChecker SG.
b.. Scanner ICC profile created using “LAB cLUT” setting on a referenced
ColorChecker SG: Low DeltaE on ColorChecker SG patches, medium deviation on
some colors outside the ColorChecker SG (red cast on some colors).
c.. Scanner ICC profile created using “Gamma Matrix” setting on a referenced
ColorChecker SG: Higher DeltaE on ColorChecker SG patches, but more respectfull
of colors outside the ColorChecker SG.
While my findings are not definitive it seem that the “Gamma Matrix” (as
suggested by Graeme) provide the smoother results and therefore the best
compromize in my opinion. “LAB cLUT” is more accurate on some colors but worst
on others. “XYZ cLUT” have strange behaviour on some colors and is therefore
unsuitable in my case.
And now some other considerations, I tought that adding more colors to the
profile (colors that are needed but are not in the ColorChecker SG) could solve
the issue allowing to build a more accurate “LAB cLUT” or even “XYZ cLUT”
profile...
I therefore wrote a tool that allowed me to add some “custom” patches to the
ti3 file generated with the ColorChecker SG using scanin.
My idea was to read the LAB values of those color patches with the same
instrument used to read the ColorChecker SG, extract the RGB values from the
scanned image and add the LAB and RGB values to the TI3 file together with the
ColorCheckerSG patches.
I therefore picked some patches from a RAL chart, trying to avoid colors
similar or already present in the ColorChecker SG. I measured those colors (LAB
with i1 Pro and RAW RGB from my scanner) in the same way I did for the
ColorChecker SG chart.
Finally I calculated a new ICC profiles with the new TI3 file (ColorChecker
SG+Custom patches).
The results clearly show some sort of uncompatibility between the ColorChecker
SG chart and the “custom patches” from the RAL chart. While the added colors
improve drammatically in the new profile, some other color, even patches on the
ColorChecker SG start showing posterizations (even in the neutral patches). The
more colors I add from the RAL chart, the more posterization I can see on the
scanned images.
I guessed the problem could be related to a different percepetion of the
ColorChecker SG and RAL charts (either in the scanner and/or
spectrophotometer). But the scanner lighting is UV free and therefore there
cannot be a FWA related issue and furthermore the scanner light spectrum is not
so far from the ideal D50 (5200°K, CRI92) and therefore I immagine metamerism
should also be low. Still there must be a reason for which the
spectrophotometer and/or the scanner does not perceive those 2 charts in the
same way !
This can be esaily verifyed measuring gray patches on both charts... the values
do not fit together for some reason.
I’m therefore struggling on the attempt of understanding:
a.. why the 2 color patches from the 2 charts cannot be used together, is
there a simple or complicated answer to this ?
b.. is there a way to determine compatibility between 2 kind of charts (i.e.
measuring metamerism level, or...) ?
c.. is there any available color chart with thousands of patches (not only
144 as in the CCSG), with a large gamut that could be used and maybe provide
more accurate results on a wider set of colors ?
Any suggestion would be appreciated.
Massimo
-----Messaggio originale-----
From: Graeme Gill
Sent: Wednesday, February 24, 2016 10:07 AM
To: argyllcms@xxxxxxxxxxxxx
Subject: [argyllcms] Re: ColProf considerations...
Massimo Colagrande wrote:
Any comment or suggestion to further improve, understand (and possibly reduce
the
processing time using the –qu option) would be appreciated.