White balance correction

Hi all,

I have hundreds of photos of fairly similar scenes (ice and
snowcovered ice), which I'm going to classify into different regions
automatically. The photoes were taken with a canon eos300D with auto
white balance. Using the camera with auto white balance was a big
mistake, and now
my photos varies from bluish to very bright.

Does anyone know how the auto white balance works on the eos300D?

The images were taken under fairly the same ligth conditions (cloudy).
Can you suggest a "smart" methode I can employ, to revert the auto
white balance setting, i.e. to make all my images look as they would
if I had used some custom white balance setting instead.


regards,
Agnar

"ahs" wrote in news:1170146508.444705.206790
@q2g2000cwa.googlegroups.com:

Hi all,

I have hundreds of photos of fairly similar scenes (ice and
snowcovered ice), which I'm going to classify into different regions
automatically. The photoes were taken with a canon eos300D with auto
white balance. Using the camera with auto white balance was a big
mistake, and now
my photos varies from bluish to very bright.

Does anyone know how the auto white balance works on the eos300D?

The images were taken under fairly the same ligth conditions (cloudy).
Can you suggest a "smart" methode I can employ, to revert the auto
white balance setting, i.e. to make all my images look as they would
if I had used some custom white balance setting instead.


regards,
Agnar




Usually, digital cameras employ simple and fast histogram-based methods for normalization or
equalization (CDF), in order to avoid unbalanced over- and under-exposed areas of the image.
Unfortunately, normalized/equalized images do not contain any inherent information about the real
exposure conditions (so that they be "reverted" afterwards). However, you can process the complete
set of images, calculate some limits and mean values (RBG) of the histogram of all the images and
then "normalize" them in the same way, so that all of them look roughly the same. That is, if the
objects, background and exposure conditions are similar (i.e. not mix daylight photos with night shots).
In any case, I would not depend much on this approach if you plan to do some analytical (algorithmic)
image analysis on them.


--
Harris