[Numpy-discussion] nd_image.affine_transform edge effects
Sat Mar 24 19:31:43 CDT 2007
> Based on my reading of the two excellent Unser papers (both the one
> that ndimage's spline code is based on, and the one that Travis
> linked to), it seems like a major point of using splines for
> interpolation is *better* behavior in the case of non-band-limited
> data than the traditional 'anti-aliasing [e.g. lowpass] filter' +
> 'resampling' + 'reconstruction [e.g. lowpass] filter' procedure.
It's certainly true that intermediate-order spline interpolants will
cause less ringing than an "ideal" sinc function. So their behaviour
is better for non-band-limited data than applying simplistic formulae
derived from the Sampling Theorem. This fact would help you out if
you don't use a low-pass filter. However, I wouldn't go as far as to
say that splines *replace* some form of low-pass filtering. I haven't
read Unser's papers in much detail, though (at least not recently)
and my applications are different from yours; it may depend exactly
what you're trying to do.
> So it seems that something is wrong if the spline interpolation
> tools in ndimage only work properly in the band-limited case, or
> require lowpass filtering.
It depends what you mean by working properly -- but in this case it
does look like something is wrong and that you figured it out in your
next post :-).
More information about the Numpy-discussion