[Insight-developers] Issues with the canny edge detection class

[Shlomo Kashani.]

Hi All,
  I am trying to use the class CannyEdgeDetectionImageFilter in an iris recognition system. One of the first steps is segmenting the iris and the pupil by and edge detector and then finding the radii of the pupil and the iris using a circular Hough Transform. I have several questions regarding the canny edge detector implementation in ITK:
   
  1-I understand that in the original paper by Canny a first derivative of a Gaussian was used while ITK uses a LoG operator (second derivative of a Gaussian). Why was this method chosen over the original suggested method? 
   
  2-in the class CannyEdgeDetectionImageFilter, I noticed that there is no way of specifying the SIZE of the Gaussian kernel that the canny edge detector would use (by a call to DiscreteGaussianImageFilter I guess ???), rather only the standard deviation can be specified and I understand that the kernel size is determined automatically from the standard deviation. Is that correct and is there a way to over ride this behavior?
   
  3-I noticed that when using ITK I have to use a standard deviation as large as *30* which would probably result in a huge Gaussian and Gaussian derivative kernel size while when using MATLAB (which uses a first derivative of a Gaussian) a standard deviation of 5 is enough. How come there is such a hug difference?
   
  Any help would be appreciated,
   
  Shlomo Kashani.  

 				
---------------------------------
Want to be your own boss? Learn how on  Yahoo! Small Business. 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.itk.org/mailman/private/insight-developers/attachments/20060625/17e77b7d/attachment.html