Hi Sander,
I found that, for my deBayer routine, leaving the outside edges at ANY 'fixed colour' gave me some really nasty 'skewed results' when it came to looking at the Histogram, etc.
In my opinion you have two 'better' choices :-
Either
a) issue a 'crop outer ring of pixels' command, or
b) perform a 'nearest neighbour interpolation' algorithm on the outer edge, and corners.
I have implemented (b) in my JSR version - it seems to work quite well. I average the three 'closest' pixels - e.g. below left, below and below right, for the missing 'top' row, and do the same thing for the other three edges. For the corners, I average the nearest three again - because I now have 'edge data' for the corners.
However, there is quite a good argument to just 'throw away' the edge data altogether (i.e. the 'crop' method) - because it is going to be noisy anyway, and because you are probably going to run a Dynamic Crop to eliminate pixel shift resulting from 'dithered' stacking (intentional, or otherwise !!)
But, and this is my point, it is nice to have 'done away' with these solid-colour edges - simply because Histograms of images WITH these edges simply 'look nasty' (even STFs are 'not nice')
That said, some users may not understand 'why' their precious images have 'shrunk' by two pixels all round.
Maybe a GUI option might be the best solution?
Cheers,