While the previous tests have shown that neither the gray world or the white world assumptions correctly estimates the illuminant under most conditions, they remain attractive due to their simplicity.
In an attempt to improve the illuminant estimate, I tried simply linearly interpolating between the illuminant estimates provided by these two assumptions (using this Isis script.)
Here are the results of "light gray" normalization, applied to the test images:
Two dimensional histograms of the corrected images are available: chromaticity and Red vs. Blue
Gamut | Illuminant A | Illuminant B | Illuminant C |
---|---|---|---|
All | |||
Little Blue |
|||
Little Red |
For quick reference, here are the test images (All gamut) :
Illuminant A | Illuminant B | Illuminant C |
---|---|---|