Color Cast Experiment Results


Here are some results of a color constancy algorithm I have been working on.

The algorithm I have been working on (labeled "After Processing") is compared to a proprietary algotithm in Photoshop and standard normalization and equalization available in GIMP.

More information is available from the links at the bottom of the page.



Original Image

After Processing

Gimp Equalize
(auto contrast had no effect)

Photoshop Auto Levels



Original Image

After Processing

Gimp Auto Contrast
(equalize result was unusable)

Photoshop Auto Levels



Original Image

After Processing

Photoshop Auto Levels

Gimp Equalize
(auto contrast had little effect)



Original Image

After Processing

Photoshop Auto Levels

Gimp Equalize
(auto contrast result was reasonable)



Original Image

After Processing

Photoshop Auto Levels

Gimp Equalize
(auto contrast had no effect)



Original Image

After Processing

Gimp Equalize
(auto contrast had no effect)

Photoshop Auto Levels



Original Image

After Processing

Gimp Auto Contrast
(equalize result was unacceptable)

Photoshop Auto Levels



Original Image

After Processing

Gimp Auto Contrast
(equalize result was not as good)

Photoshop Auto Levels


Click here for a lot more results...


Click here for results on real world data...


Click here for research info...


Click here for results on some of the SFU data...


Click here for the most recent results on all of the SFU data...


Misc other results pages: 1


The results on this page were updated on 11/28/00.

The relevant paramters are:

setenv ih_fname ill_tc5e_cp21_sm01_more69.hist
setenv sh_fname surfb_tc5e_cp21_sm01_sm00_more69.hist
setenv cc "ccolor -hw 1 -ra 0 -cp 86 -hc 1 -nm 1 -st 0 -pw 0.0 -et 0 -g 2.2 -ih $bhp/$ih_fname -sh $bhp/$sh_fname -l log.txt"
The histogram is the first cycle histogram starting with uncorrected images.

No brightness or contrast adjustments are made.


Charles Rosenberg
Last modified: Tue Nov 28 21:28:11 EST 2000