Re: [OctDev] imrotate

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

Paul Kienzle <pki...@us...> wrote on Tue, 22 Feb 2005
01:09:38 -0500:

> FWIW, matlab also exhibits the alternating bright-dim stripes but less
> severely than yours.

For imrotate(eye(...), ..., ...) I just verified that for all of
"nearest", "bilinear" and "bicubic", Matlab's (6.5) and my code
produce *exactly* the same result, modulo the following:

- My code crops the image a little more aggressively than Matlab
  does. I round the size such that 90.1 and 90 degrees (e.g.) give the
  same result. Matlab uses the ceil. It's a matter of taste.

  Moreover, Matlab appears to pad the image with zeros, while I do not
  pad. Thus, Matlab fills in more values/produces a larger
  image. However, I do not see why we should pad; the value of zero is
  about as arbitrary as any other value. My code fills in (the
  arbitrary value of) zero wherever the result is not defined. Again,
  it's a matter of taste, but my method yields simpler code.

- Both Matlab and my code may produce an odd-size result image for an
  even-size source image, and vice versa. In those cases where
  Matlab's and my code produce resulting image sizes of opposing
  parity, the results are obviously not identical, but neither is more
  "correct" than the other.

Justus

--=20
Justus H. Piater, Ph.D.         http://www.montefiore.ulg.ac.be/~piater/
Institut Montefiore, B28        Phone: +32-4-366-2279
Universit=E9 de Li=E8ge, Belgium    Fax:   +32-4-366-2620