[Octave-bug-tracker] [bug #44799] image package: imrotate chrashes when using 'spline' as interpolation method

[Hartmut]

Follow-up Comment #6, bug #44799 (project octave):

As long as this (Octave core) issue is not fixed, should we maybe change the
documentation in the image package accordingly?

Currently the help string of imrotate.m reads:

The optional argument METHOD defines the interpolation method to be used.  All
methods supported by 'interp2' can be used.


And the help string of interp2.m (in Octave core) then claims:

The interpolation METHOD is one of:
[...]
"spline"
[...]


This leads users of imrotate to the wrong conclusion, that "spline" is a valid
option for the method parameter of imrotate. But imrotate then only throws an
(interp2) error, since a rotated pixel grid is nearly never evenly spaced, as
currently needed by interp2 (more precise: needed by __splinen__.m).

So if (core) interp2.m isn't fixed in the near future, I would suggest to add
the text "(except for 'spline')" to the help string of (image package)
imrotate. This help text would then read like this:

The optional argument METHOD defines the interpolation method to be used.  All
methods supported by 'interp2' can be used (except for 'spline').  In
addition, Fourier interpolation by decomposing the rotation matrix into 3
shears can be used with the 'fourier' method.  By default, the 'nearest'
method is used.

For MATLAB compatibility, the methods 'bicubic' (same as 'cubic'), 'bilinear'
and 'triangle' (both the same as 'linear') are also supported.



    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?44799>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/