[Octave-bug-tracker] [bug #46099] image package: edge(I, 'canny') gives very bad quality results

[Carnë Draug]

Follow-up Comment #15, bug #46099 (project octave):

I have pushed your changes with your tests (I made some changes to the tests
and included the Matlab threshold values).

http://hg.code.sf.net/p/octave/image/rev/cb24795f0d8f

Your tests seemed to be Matlab compatible (passes both your tests) but you
claim there's some differences (could you add such a test?). Maybe it's
because of the way it guesses the threshold values?

One thing I noticed was that you explicitly filter the image once per
dimension but did not explain why not use a 2d filter.  I have this patch
which seems to not break anything:


-      ## Gaussian filtering to change the edge scale (treat dimensions
seperatly)
+      ## Gaussian filtering to change the edge scale
       len = 8 * ceil (sigma);
-      x = -0.5 * (len-1) : 0.5 * (len-1);
-      gauss = exp (-(x.^2)/(2*sigma^2));
-      gauss = gauss ./ sum (gauss);
-      J = imfilter (im, gauss, "replicate");
-      J = imfilter (J, gauss', "replicate");
+      J = imfilter (im, fspecial ("gaussian", [len len], sigma),
"replicate");


Did I miss anything that I should not push it?

> @Carne: Before I forget it: The results of the Canny edge detector are
supposed to be 4-connected contours, not 8-connected contours. Maybe this can
explain your differences when using imreconstruct instead of nonmax_suppress.

Looking into nonmax_supress it seems to be looking in the diagonal values (45
and 135 degrees), so I'm assuming it is connectivity 8.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?46099>

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