[SciPy-User] possible bug in scipy.ndimage.interpolation.shift

[Gary Ruben]

I've just used scipy.ndimage.interpolation.shift to align some greyscale 
images represented as float arrays to sub-pixel accuracy which generally 
works well. However, I discovered that the resulting image array 
sometimes contained some negative values. I think there were about 7 or 
8 so they are probably occurring near the corners. For the moment I just 
forced any -ve values to 0 as a workaround. I don't feel that I have 
time to produce a minimal example at the moment because I'm writing a 
thesis but I thought it better to flag it here than keep silent. I'll 
keep this message in my inbox and when I'm finished in a month or so 
I'll raise a ticket.

Gary