[argyllcms] Re: White-point scaling for input device profiles, parameter - U
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Wed, 18 Jan 2012 14:11:54 +1100
Stephen T wrote:
> The white and black points look OK (no impossible colours, e.g. Y > 1, XYZ <
> 0)
> but I find that the "default" profile delivers better-looking images than the
> alternate profile with scaled white point.
In what way does it look better ?
Chromatically the result shouldn't change, it just should be scaled
linearly in XYZ space.
> P.S. The raw photo converter I'm using performs white balance (suboptimally,
> I know) on device RGB.
Keep in mind that few programs use just the input device profile and then work
in an XYZ working space, so probably the input profile is being linked
with an RGB output profile. Setting a -U factor > 1 will scale the apparent
gamut size down, and while setting a white point afterwards should
scale it up again, the result of doing this in a gamma corrected RGB
space will be different to doing it in the linear light XYZ space
that would be the true inverse of the effect of -U.
[ Ideally the raw photo converter would do the white point selection
using a chromatic transform in XYZ space. ]
(The real aim is to give some more headroom above that implied
by the test chart).
Graeme Gill.
Other related posts:
