[OctDev] New imrotate: nearest, bilinear, bicubic, Fourier

[Justus P. <Jus...@UL...>]

Hi,

Here's a new version of imrotate() for review and inclusion in
octave-forge.

Here are the changes with respect to the previous version:

- Integration of Jeff Orchard's original Fourier-based method into the
  same interface by mapping to his original file. The latter should be
  hidden from the user in some way (are there private/ directories in
  Octave as there are in Matlab?).

- Addition of bicubic interpolation. This code is not limited to this
  particular use and could perhaps be moved into a separate file. The
  source data must be a matrix, and the interpolated points can be
  anywhere within that grid (in some way, this is about half way
  between interp2() and griddata()).

- My previous version rounded the target intensities to integer
  values. This is no longer done. Also, bicubically interpolated gray
  values may overshoot and thus exceed the [0,256) range. The same
  holds for the Fourier method.

I attach a comprehensive test suite. Run "testimrotate".
The files that should go into CVS are imrotate.m and imrotate_Fourier.m.

Cheers,
Justus