Winsor Crosby wrote:
> Is that really what it does? That explains a lot about why it does not work.
>
Or, put the other way 'round, the fact that it doesn't work points to a
simplistic operation.
The examples in my little test come form my own effort to understand how
Auto WB works in practice. The PS example is simply the image as it came
from the scanner, using the icc profile for the film, before and after
application of the Auto-Color tool in PS. As you can see, it turns the
green background to gray.
As to "Is that really what it does?", what other choice is there? Given
no knowledge of content and no neutral reference point, what can a
programmer do? Absent any other information, the only assumption that
makes any sense at all is that of an overall neutral balance of all
colors in the image.
One may make some tweaks, and I think PS does this, and it likely
accounts for the differences in auto WB in various cameras and software
apps. One improvement might be to exempt individual small areas of color
very different from the overall image from strong correction. Notice how
relatively little the green of the clover foliage and the red of the
flower are changed, compared to the vast change of the door color. I
think the Auto-Color is actually pretty smart. It just doesn't have a
chance on an image like this, where all assumptions about neutrality fail.
Another enhancement could be to handle the blue areas of images with
lots of blue in the top half differently from the rest of the image.
Somebody (Nikon?) has analyzed millions(?) of images to determine
recognition patterns to allow customized processing. That would
certainly allow custom WB treatment, as well.
Moose
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|