An Edge Detection Method by Combining Fuzzy Logic and Neural Network

Abstract

An edge detection method by combining fuzzy logic and neural network is proposed in this paper. First, the distance measures between the feature vector in 4 directions and the six edge prototype vectors for each pixel are taken as input pattern and fed into input layer of the self-organizing competitive neural network. Classifying the type of edge through this network, the thick edge image is obtained. After classification, we utilize the competitive rule to thin the thick edge image in order to get the fine edge image. Finally, the speckle edges are discarded from the edge image, thus the final optimal edge image is got. We compared the edge images obtained from our method with that from Canny’s one and Sobel’s one in our experiments. The experimental results show that the effect of our method is superior to other two methods and the robusticity of our method is better.