[argyllcms] Re: Why colprof doesn't generate identity profile for synthetic image?

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 03 Sep 2012 12:34:35 +1000

Mikhail Matrosov wrote:
> Yes, that works. But I thought I need full power of cLUT to calibrate a
> camera?


If you are going to use a model with a large number of free variables,
then you need to set them using a large number of test patches.
24 are not enough.

Argyll's cLUT's do have a smoothness constraint, but this doesn't
make unconstrained cLUT points conform to an additive model.

> How is "single gamma+matrix" enough?

It's enough if it matches the behaviour of your device. Perfectly
additive devices can be modelled perfectly by curves + matrix.

Graeme Gill.

Other related posts: