A fuzzy approach to edge detection and representation

This paper proposes a method of edge representation which is more general than the conventional edge representation method. While the localization of conventional edges is considered as the problem of determining its existence at a certain location, that is, either 0 (non-edge) or 1 (an edge point), our representation method defines the location as some measure of plausibility of edges using the fuzzy set theory. Thus, if the location of intensity edges is unclear, or if the intensity function is different from ideal edge models, the degree of edgeness and location is represented as a fuzzy membership value. The conventional edge is a special case of our representation-a case where the input intensity is almost same as the model of edges (for example, a step function) with little degradation or noise. We show a computation model of our fuzzy edge representation method followed by experimental results.