<div class="gmail_quote">On Thu, Mar 12, 2009 at 4:35 PM, Wen Shi <span dir="ltr"><<a href="mailto:wsa18@sfu.ca">wsa18@sfu.ca</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi Ghassan and guys,<br>
<br>
As we discussed before, for the cannyedgedetectionfilter, since it deals with noise by first convolve the image with a Guassian filter,<br>
I supposed that maybe the data from the output filter can be a good measure for the edges.<br>
But when I tried a brain image with the filter, I saw lots of noisy edges in the processed image.<br>
Even in the region in which the original image has nothing(no edge clues there) there, the processed image has a lot of short bright edges.</blockquote><div><br>Try any of the following<br>
<br>- Increase the hysteresis / strength thresholds on the canny filter, so it picks up the stronger edges.<br>- Increase the sigma being used by the canny filter, so it picks up the edges at a larger scale.<br>- Smooth the images prior to canny, perhaps with anisotropic diffusion etc.<br>
<br>--<br>karthik<br></div></div><br>