Topologic edge operators

ABSTRACT PhD Dissertation TOPOLOGIC EDGE OPERATORS ELİF AYBAR Anadolu University Graduate School of Sciences Electrical and Electronics Engineering Program Supervisor: Prof.Dr.Atila BARKANA 2003, 78 pages In this thesis, two new edge detection methods that make use of topological properties are proposed. Connectivity map (conn-map) is a quantitative depiction of how the center pixel is connected to others in an observation window. Both edge detection methods compute connectivity map in a sliding window and use the connectivity information. First method use an self adaptive algorithm to produce an output. The second applies Sobel edge operator on the conn-map. For simplicity purposes, Sobel edge operator of size 3x3 has been generally used in edge detection operations. In the literature, in general, larger Sobel operators (e.g. 5x5, 7x7and more) are not utilized due to the reason that they produce thicker edges since the number of edges included in larger windows increases. Connectivity map filters the edges that are not connected to the center pixel, therefore, both methods yield leaner edges. Performances in noisy image environment are also studied. Especially, the second method is determined to have a superior performance compared to conventional Sobel operator. Keywords: Edge Detection, Edge Operators, Image Topology, Connectedness Map, Sobel Operator