On Wed, 25 Apr 2007, Christopher Friedt wrote:
This is probably a very common question to ask, but can anyone explain
vaguely, or point me to a better description, of how one could
automatically optimize the image so that it appears 'natural'?
What I do is select a part of the image to focus on, then compute the
average U and average V of that selection, and then make PWC calls to set
the white balance until the average is quite close to 0.
For example, I've been selecting everything for Y>100 for example, if
viewers are going to look at bright objects rather than dark ones. If
edges and intricate surfaces are more interesting, then use a high-pass
filter (edge detector) to decide whether a pixel is in or out.
Instead of making a clear-cut selection, you may also give weights to
pixels, e.g. proportionally to Y, to Y*(256-Y), or the norm the output of
a high-pass filter. (the norm being square root of Y*Y+U*U+V*V, perhaps
with special weights for U and V (?))
Does this have something to do with gamma values?
Gamma is one of many factors that you may have to play with. AFAIK it's
hard for a program to self-optimise its vision without optimising a
certain monitor-camera pair and not any of those separately.
_ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard - tél:+1.514.383.3801, Montréal QC Canada _______________________________________________
pwc mailing list
http://lists.saillard.org/mailman/listinfo/pwc
|