New reason for XYZ

[Kai-Uwe Behrmann]

Am 07.03.07, 11:01 +0100 schrieb Cyrille Berger:

> > > > Can you display XYZ correctly, without banding? Try gradients. I
> > > > remember about problems and did not continue.
> > >
> > > I am not sure I understand what you mean, but for a gradient I get this
> > > http://cyrille.diwi.org/tmp/krita/screenshot12.png.
> >
> > .. and a gray scale gradient?
> ah I see... But there is obviously something wrong, as if I draw my gradient 
> in RGB8 then convert to XYZ16 I also have the blueish halo. It's more a 
> problem from converting RGB8 to XYZ16 I think. But that's really curious, I 
> would think XYZ to have less "problems" than any other colorspace...

The difficulty arrises by the usable points in the conversion LUT. As XYZ 
is very huge, the used point in the transforming LUT to tell about the 
right transistion are a few only. With fewer points less accuracy ...

To visualise look at the devicelink tables in ICC Examin.

regards
Kai-Uwe Behrmann
--
development for color management 
www.behrmann.name + www.oyranos.org + www.cinepaint.org