I was playing with the despeckle option in the grayscale tab of ufraw. If I, e.g., select to view the red channel, then use a window size of 30 and 5 repetitions, I can see the scale of the displayed image change (the outer image parts move further out). If I do that for the blue channel, the image shrinks (outer parts move inside). So I wonder wether this is an error, or intended:
The image I'm processing is a raw image of a Lumix DMC-LX3, which does have quite some color error, the blue image is substantially larger magnified than the others (I need to set the kb linear color factor to 1.002 to compensate). My impression is the algorithm somehow uses the whole image information of the input image (i.e., averages all channels) for some reference? Or why is it working different for different channels (and why is it changing the scale on first hand)?

This does not occur with images that do not have a (substantial) TCA. Can someone confirm this? Should I file a bug report?

I'm working with the cvs version of ufraw (checkout Oct 17)