[argyllcms] Re: Profiling in-camera color transformation

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sun, 16 Nov 2008 02:38:34 +1100

Pascal de Bruijn wrote:

Yes, but not in the normal way. Normally profiling a camera would mean you're
making a profile faithfully reproduce accurate colors.

Device profiles don't reproduce color, they characterize a device.
Naturally the idea is that the characteristics of the device
be faithfully recorded.

You can use a profile to reproduce color by linking profiles.
The CIE values (Profile Connection Space) is the common
element that allows two profiles to be linked.

When you link profiles you can choose what sort of
rendering intent you want, which may be aimed at
colorimetric reproduction, or you may choose a perceptual
(preferred) reproduction.

My camera, does internally apply a color conversion. Which I can't
seem to be able
to reproduce. So I'm trying to abuse ICC profiles for that.

I want to "profile" the difference, between my raw (developed
liniearly), and the JPG
produced by my camera.

You seem to be wanting to create a device to device link directly from
measured data then. Argyll's colprof won't allow you to do this,
although there may be some tricks that will work.

But is there a way to make this work?

The only thing that crosses my mind is to convert one of the
RGB spaces into CIE using a known profile such as sRGB, create
the device "profile", and then link it colorimetrically with the sRGB
profile to form a device to device link. Probably the JPEG
RGB values are the thing to convert using sRGB. You may
have to think about where to set absolute and relative colorimetric
in the fakeread and collink (ie. make it all absolute ??).

You would have to get the IMG_4438_JPG.val file and lookup
each RGB value. You can do this with some messing around
using fakeread I think. You may have to hand edit the CGATS
headers a bit to turn the .val into a .ti2.

Graeme Gill.

Other related posts: