[argyllcms] Looping back a target

  • From: Joe Moore <jpvlsmv@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 12 Sep 2011 18:08:15 -0400

I am new to color management, so please bear with me if I'm asking
dumb questions.  I have noticed that pictures scanned in from my
scanner (HP Scanjet 5590) end up with radically different colors when
displayed on screen, than they do on paper.  I assume that I can
characterize the scanner, then use that information to remap the
scanned images into something that looks right on screen.  I'm not
looking for something incredibly precise, just to have flesh-toned
skin.  The Autocorrect settings in the scanner software, well, don't.

I've created a test target with targen and printtarg to a TIFF file.
But before I took this into the physical world, I wanted to see how
the rest of the image processing would go.  My overall plan is to take
the target down to my local photo print place and get something that's
"close enough" to what's on the screen that I'm satisfied.

targen -v -d 2 target4
printtarg -i SS -v -a .4 -t 300 -p 4x6 -s -m 10 target4
scanin -dipn -v1 target4.tif target4.cht target4.ti2 diag.tif

The target I've created (through the various steps) can be found at

When I feed the target.tif back into colprof, however (which I think
should create a null ICC profile, i.e. one that makes no changes to
the color space, however looking at the curves in the ICC Profile
Inspector, they are not flat at all.

Looking at the TI3 file, the deviations are very small (<10^-13).  But
colprof -v has a very interesting idea of what "White" is in this

target4>colprof -v target4
No of test patches = 836
Estimating white point
Picked white patch 283 with XYZ = 0.907340 0.982510 0.858160, Lab =
99.319739 -7.096156 -3.825529
Picked black patch 5 with XYZ = 0.010000 0.010000 0.010000, Lab =
8.991442 1.317040 -2.855386
Approximate White point XYZ = 0.907340 0.982510 0.858160, Lab =
99.319739 -7.096156 -3.825529

As you can see from the RGB values it picked, it is nowhere near a
pure color (the choice of black is a lot closer to black).  Guessing
from the lines in target4.TI3, it's looking at patch V06 which is RGB

Am I missing something in this process?  Is it a futile quest to try
to get a 1:1 ICC profile out?  Why is colprof not going with the true
white patch at R10, R11, Q19, or R12?


Other related posts: