Am 24.11.2012 16:42, schrieb Tom Schumm:
and i1Profiler does generate a reasonably accurate and usable profile, and it at least claims to be using a matrix profile.
How did you assess that is accurate? Did you measure a large number (say 1000) of color samples and compare them against the profile with profcheck? What average/max error did you get this way?
Is there something I could be doing wrong that would let me create an accurate LUT profile but a dramatically wrong matrix profile with ArgyllCMS?
Provided that there are no unusual repeatability and/or measurement errors, this is likely just the indication that the display isn't as additive as you might believe (at least in its current operating mode). [ In order to verify additivity directly, you would need to measure the display with a spectrometer. ]
I can add some additional information. After playing around with it for a while, I'm sure the Argyll profile is the one that is wrong - it has very high Delta E for some colors, and subjectively looks quite wrong (reds are really really orange, not a little bit orange).
What is the average delta E of the matrix profile, as reported by colprof? (average 3, max 10 can be IMO still reasonable - one of my displays isn't better either) Note that colprof's priority is reducing the _average_ squared error of all color samples given in the .ti3 file, at the cost of larger individual errors for some colors samples. So your large error at red could well be the price you have to pay in order to achieve a better average error. Best Regards, Gerhard