[argyllcms] Re: Profile input white not mapping to output white

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Tue, 27 Nov 2012 23:23:16 +1100

Gerhard Fuernkranz wrote:
> For camera profiling, it is IMO not so clear that the color and luminance of 
> the white
> patch on a test chart has any specific meaning for the scenes which are 
> eventually
> captured with the camera (particularly if the white patch does not even 
> happen to be
> spectrally flat, as it is the case for some targets).


        there's nothing else there to set a relative colorimetric white point
though, and for situations where it is not relevant, then I would
imagine some other workflow using absolute colorimetric would be
the right approach (ie. a manual or auto white point/white balance selection.)

> I'd possibly be more liberal and
> give the users the choice to select the profile's media white point location 
> freely,
> depending on their particular shooting scenario and/or workflow, in order 
> that applying
> the profile with relative colorimetric intent will eventually clip at the 
> desired white
> point.

I'll have to think about that.

> [ This still does not solve any extrapolation issues of course, unless the 
> chosen WP is
> inside the gamut of the target. Btw, which model do you use internally for 
> extrapolation?
> -am, -ag, -aG ? ]

It's a special internal matrix model which I tried to tune up to be reasonable
with the couple of examples I have. It uses gamma per channel curves with one
extra shaper order + matrix, with some weighting towards the neutral
patch samples, the idea being to reduce the influence of chromatic
samples on the neutral axis response.


Other related posts: