Ahmad Keshavarz

The subject of this thesis is to propose a robust edge detector based on the use of anisotropic Gaussian kernels. Because of the high importance of edges in image processing, there exists many edge detection methods. Among them, Canny edge detection method is most important. It is shown that the seminal Canny edge detector may miss some obvious crossing edge details and the presented method overcomes these difficulties. We begin with the use of anisotropic Gaussian kernel for noise suppression and highlighting edges. Then, it is shown that why seminal Canny edge detector may miss some obvious edges. Afterwards, based up on the use of anisotropic Gaussian kernel for noise suppression, a revised edge extraction method is proposed. The experiment results show that the proposed algorithm can obtain better performance for noise-free and noisy images.