A fuzzy if-then approach to edge detection

An edge detection approach based on fuzzy if-then rules is presented. This method avoids the difficulties of selecting parameter values in most of the edge detectors when no information about the images is known in advance. Combining all the if-then rules generates a set of potential edge pixels. The membership value of being an edge point for each pixel is assigned by a membership function. The pseudocentroid of a set of potential edge points is used as the threshold for the decision of selecting a real set of edge pixels. Comparison studies with the gradient, Laplacian, and Laplacian of Gaussian edge detectors having fixed parameters are provided. The empirical results show that the edge detector based on fuzzy if-then rules is generally more applicable to a wider class of images ranging from clear to very vague images.<>