contrib/oxl/osl/osl_canny_rothwell.h File Reference

A class for performing traditional Canny edge detection. More...

#include <osl/osl_canny_base.h>
#include <vil1/vil1_image.h>

Go to the source code of this file.

Classes
class	osl_canny_rothwell

---

Detailed Description

A class for performing traditional Canny edge detection.

The basic implementation is as close to that described in Canny's thesis as possible, though sub-pixel interpolation of the final edge output has been included, as has edge thinning. Note that there are probably differences between this implementation and the documentation (that is his Master's thesis).

The improved implementation includes the patching of single pixel gaps in the Edgel chains, as well as improved search for edgels near to edgelchain terminations (dangling ends). At these places the scale of the smoothing convolution kernel is reduced dynamically to remove interference from corners (if they exist) near to the ends. Overall, this yields a much better understanding of the image topology.

Note that nothing special has been done around the border of the image; we have simply ignored a border of size width_ all the way round. Perhaps this should be changed to provide consistency with the rest of TargetJr.

Author:

Charlie Rothwell - 25/1/92 GE Corporate Research and Development

    Modifications
     Samer Abdallah - 5/10/95
     Robotics Research Group, Oxford University
      Most members and methods placed in new base class osl_canny_rothwellBase.
      Fixed bugs in Final_hysteresis() and Jump_single_pixels().
   

Definition in file osl_canny_rothwell.h.