Hi there,
I’m quite literally trying to calibrate a scanner so I can pretend it’s a
colorimeter because my samples do not have homogeneous surfaces that are large
enough for a proper spectrophotometer.
I wonder what is the best one can get in calibrating a scanner? For most
samples that are non-pathological, is it ever going to be possible to achieve,
say, DE less than 2 or so?
So far I have tried to mess with an Epson Perfection V550 Photo and the Wolf
Faust “affordable” IT8 target. Doing “colprof -qh -ax -u” the profile check
says peak err = 2.156632, avg err = 0.367140 which looks alright. But in
reality for most samples I have tried, the spectrophotometer reading and the
scanner output still disagree quite substantially usually with DE 5 or even
worse.
I presume I need a better target or at least something with bigger patches so I
can actually measure them with a spectrophotometer because obviously the XYZ
data from the manufacturer won’t be too accurate? I’m thinking if the Xrite
ColorChecker SG is going to solve the problem? Or perhaps can I get something
like a Pantone or RAL colour chart (they have more colours) and create my own
.cht file? Or even creating my own target if that’s going to help?
Besides, as an outsider, it’s still quite difficult for me to comprehend why on
earth can’t the icc profile, at least in cLUT mode, map the patches to
precisely the XYZ values in the text file? In other words when you are doing
the interpolation you don’t know how the device actually behave so you have to
guess (or not bother at all and *assume* everything is linear...), but the
patches in the IT8 target has XYZ values already given in the text file so why
do we map them to anything else? I presume there must be a very good reason but
why?
Many thanks,
Zi Wang
