[Digikam-devel] [Bug 236304] Add an button in the Whitebalance Settings to load Camera WB

Alex Tutubalin lexa at lexa.ru
Thu May 6 19:32:18 BST 2010


https://bugs.kde.org/show_bug.cgi?id=236304





--- Comment #6 from Alex Tutubalin <lexa lexa ru>  2010-05-06 20:32:15 ---
Jens,

LibRaw uses 3 (or 4 for two-different-greens cameras) scale coefficients for
white balance. These coefficients are not camera independed, but includes both
color balance and chanel sensitivity adjustment in one step. These coefficients
are stored in Exif data (for 'as shot' balance), so it is straightforward way.

It is impossible to separate sensitivity part from 'color temperature/tint'
part without experiments with *each* supported camera (300+). You need test
shot of good gray card under well-controlled light (5000K or so). I have no
such data.
Also, this is not good way because original data (provided by camera) are 3-4
coefficients (green one is 1.0 for many cameras).

Adobe uses another way, similar for von Kries transformation. This way is
theoretically right (althought, color from 'Raw Photo Processor' which uses
4-coefficients way is way better than from Adobe's software), but in practice
it leads to more noise because weak noisy channels are mixed with strong one.

Summary:
 1) LibRaw uses 3(4) coefficients way, so any calling application should
provide these data to postprocessing stage.
 2) Cameras store these coefficients in Exif data, so it is easy to provide
this coefficients to LibRaw (and LibRaw itself reads them).
 3) If you want to play with Von Kries/Bradford white point adjustments, it is
up to you, but starting point is camera's coefficients and you should provide
same (may be, modified) coefficients to LibRaw postprocessing.

Also, I see no way to deal with two-different-green cameras and von
Kries/Bradford transformation.

-- 
Configure bugmail: https://bugs.kde.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.



More information about the Digikam-devel mailing list