[argyllcms] Re: Profiling between two images

  • From: Pascal de Bruijn <pmjdebruijn@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 19 Feb 2009 08:33:54 +0100

On Tue, Feb 17, 2009 at 7:16 PM, Gerhard Fuernkranz <nospam456@xxxxxx> wrote:
> Pascal de Bruijn wrote:
>> Some loss in mathematical accuracy is very acceptable...
>> However, can fiddling with these values also affect the generated matrix? Or
>> will it solely affect the shaper?
> Matrix and shapers are optimized together, minimizing the sum of the
> error term plus the regularization term, thus I would not be surprised
> if in conjunction with the changed shapers a different matrix would turn
> out to yield smaller errors.

Initially a was afraid of this... However, in practise the matrix does
not seems to change

With pretty extreme values in xicc/xmatric.c all my curves are very smooth:

#define XSHAPE_BASE  0.001
#define XSHAPE_HBASE 0.01

Accuracy is indeed a bit reduced, but considering the fact that I'm
trying to proceduce
"generic" profiles, this is very acceptable, especially since the
profiles seem to be
giving me much better results in gradient area's.

Pascal de Bruijn

Other related posts: