Monthly Archives: September 2010

Twitter Weekly Updates for 2010-09-26

Local Maxima in Edge Detection

I’ve been working on some homework for computer vision. It’s such an interesting class because of how rich the visualization of my code results can be. Matlab has also been a real pleasure to work with. Everything is well documented and the IDE has features I haven’t seen in other IDEs. I am working on an assignment in edge detection.

There are several steps required including performing a convolution of the image with with the derivative of the Gaussian, determining the magnitude of the gradient, and using it to find local maxima (in terms of intensity) of the image. That’s the part I’m at right now. Instead of using the x and y components of the gradient to compute edge orientation, I just determine local maxima by querying all the pixel neighbors of each pixel. This was quicker and helped me to make sure my code worked. Here’s the magnitude of the gradient of the image and the local maxima as computed by my algorithm. They’re very bad points for edge detection, but I’m happy I get something resembling the shapes of the image! One step at a time.

Twitter Weekly Updates for 2010-09-19

Matlab’s Image Edge Detection

I’ve been doing some homework for Computer Vision class. Matlab apparently has extensive image processing capabilities. One handy function is “edge” which will return a zero-filled matrix with 1s corresponding to edges.

The picture shows the results of using various types of edge detection.

From left to right, top to bottom, the methods of edge detection are: Sobel, Prewitt, Laplacian of Gaussian, Canny, Roberts, and Zero-cross