Associative approach for edge detection

An algorithm for edge detection applying the Associative approach is presented in this paper. An autoassociative memory is built from the original image. Nine eigenvectors are obtained from that matrix, then an eigenvector is selected and used it as a mask together with its transpose, both masks are convolved with the original image and added; the result is the detection of the edges. We compare our proposal with the most common edge detection algorithms as Canny, Prewitt, Sobel and Roberts. The comparison shows that we obtain similar results as Roberts algorithm, and when the image is has high frequencies, Alpha-Beta edge detector results are very similar than the other four algorithms.