Canny Edge Detector with Half Entropy

An adaptive Canny edge detector for gray scale images has been presented in this paper. Initially, gradient of gray scale images were obtained by means of Sobel operator and the entropy of gradient images were calculated. Then the gray level, in which the half of entropy of gradient image is achieved, is selected as the upper threshold of Canny edge detector. Additionally the upper and lower thresholds were also selected by means of median, mean, Kapur's maximum entropy and Otsu's between-class variance method and the results have been comparatively discussed.