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

  • From: Iliah Borg <ib@xxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 26 Nov 2012 12:21:26 -0500

Dear Gerhard,

> one of the problems with applying the profile after white-balancing is of 
> course that the profile is not applied to the full raw RGB cube

I guess it is better to take into account how CFA is designed and what are the 
intended mechanisms to deal with raw data. One of the things is - it is RG1G2B, 
and even if G1 filter has the same colour as G2 they still are often rotated 
slightly differently to minimize optical cross-talk and thus have different 
efficiency. Sometimes we deal with non-RGB CFAs; and multi-spectrum sensors are 
coming too (5 to 7 different types of filters in CFA).

The "body" of the capture for RGGB-type Bayer is asymmetrical for a normal 
un-filtered exposure. In overwhelming majority of cases spectral transmissions 
of R, G, and B channels are not independent and overlap, and what looks like an 
RGB cube is not RGB (hard to call yellow green) and not a perceptual cube. 
Making 4 or 3 exposures for each channel to reach "to the right" condition and 
recombining to get rid of white balance results in essentially the same profile 
as from one exposure normalized for a green channel to nearly reach the 
saturation point and with white balance applied (only the dispersion is less 
with recombination).

Iliah Borg

Other related posts: