[argyllcms] Re: how to verify measured target values
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sat, 11 Nov 2006 16:42:41 +1100
Mike Russell wrote:
I'm getting a max deltaE of about 4 from the profile command.
That seems pretty reasonable then.
Yet my inkjet prints based on this profile are very washed out and yellow,
> due to red and green being added at relatively high levels in the shadow,
> starting at RGB(0,0,0). Blue does not show up in the shadow until about
Lab(15,0,0).
Would you have a recommendation on where I should start looking first for
the problem?
Well there are many possible approaches:
Examine the profile and other elements involved in the reproduction,
(such as system and application settings, source profile) and see if
you can recognize what the problem is.
Test each element involved in the reproduction. Print some test chart
device value through the test chart print path and read them, and
check against the .ti3 values. Convert a test image to device values
using cctiff to do the link and conversion, and print it via the same
settings used to print the test chart. See if the image looks the same
as that printed with the system and applications using the profiles.
Convert the device space test image back through a proofing transform
(ie. device profile colorimetric to your display colorimetric),
and look at the resulting image on your display. For a problem color,
lookup what PCS value it has in the input profile (icclu). Look that
PCS up in the device profile to get the device values. Look those
device values up in the device profile (absolute colorimetric). Measure
that problem color on the printout, and compare to the calculated value.
etc.
Re-do the profile, carefully checking that it's being done with the
right system and applications settings at each step, and see if anything
improves.
The most tedious but sure fire method is the middle one. You just need
to invent a way of checking each step, to see which step is at fault.
hope this helps....
Graeme Gill.
Other related posts:
