Re: [OctDev] imrotate

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

Paul,

Paul Kienzle <pki...@us...> wrote on Sat, 19 Feb 2005
15:47:00 -0500:

> Here are some more tests you might include:
>
>    % Make sure the dithering preserves pixel density
>    t=3Dzeros(1,5);
>    for n=3D1:15; t(n) =3D sum(imrotate(eye(n),20,'bilinear')(:)); end
>    assert(t,[1:15]);

This test in fact works. The only difference is that my code always
produces a slightly smaller sum because it does not pad the image (see
my previous message for a brief discussion). With growing image sizes
and fixed rotation angle, the difference is bounded:
assert(t,[1:n], 2.2) should pass for any n.

>    % Make sure the size matches the expected size assuming square pixels.
>    angle=3D20; t=3Dzeros(15,2);
>    for n=3D1:15; t(n,:) =3D size(imrotate(ones(n,3*n),angle,'bilinear'));=
 end
>    a =3D angle*pi/180;
>    x =3D ceil([(1:n);3*(1:n)]'*[cos(a) sin(a);sin(a) cos(a)]);
>    assert(t,x);

If you insist... the code for computing the image size is trivial. But
to be fair, use round() because this is what imrotate does. In that
case, a tolerance of 1 is sufficient to pass the assert(), except for
the very first element. But I really would not include this test. It
is only a rough estimate, the code in imrotate is at least as simple
and does it better.

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