[argyllcms] Re: Collink and device link profiles

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 08 Jul 2011 18:47:23 +1000

Michael Schulz wrote:
thank you for the reply and for the indication how to check the result with
cctiff. In the meantime I did more trials with collink's device link option.
What I found out is, there is a slight yellow colour cast in the converted
images (rgb2cmyk and cmyk2cmyk) - I misunderstood this as an over
saturation. It becomes visible in skin tones and noticeably if one compares
the yellow colour channels of images that have been converted across the PCS
with device link conversions.

        it's not possible to say why you see this effect without
being able to exactly duplicate your situation, including devices and
viewing environment. I'm not aware of any tendency of Argyll device
links to have any color cast - they merely do what the profiles tell
them to do. Different intents and profiles will interact in different ways.

I don't understand where it comes from but I assume that it might be the
black ink generation and not the gamut mapping. I tried -ke, -kt, -kx, -kr
and -kh; beside of the yellow colour cast -kt gives the best results. Does
it makes any sense trying a black ink generation with the -kp... flag?

It's unlikely to be black ink settings because it's "closed loop" - unless
you are outside the devices gamut, you always get an accurate color value as 
the device profile can describe the device behaviour. If it does appear to
have an effect in a real world viewing situation then you may be experiencing
metameric failure due to your illuminant not matching what you've told
Argyll. (Note that by default you get D50 spectrum, the ICC standard, and
there are almost no real D50 spectrum viewing booths).

You aren't choosing a particular black generation if you are using -kt -
you are letting the incoming CMYK choose it. If you want to choose a particular
black generation then yes, you need to use -kp and choose appropriate
parameters, as outline in the tutorial:

What I don't understand in your answer is the last paragraph - why do one
gets multiple gamut mapping transformations? I thought with device link
profiles there is only one B2A table used?

But if you were to use a perceptual A2B table to "preview" the
result of a device link transform, then you may be applying a reverse
gamut mapping in the process (depending on what software made the
device profile). This can result in an over saturated looking result.

Graeme Gill.

Other related posts: