An Improved Algorithm for Canny Edge Detection with Adaptive Threshold

To overcome the difficulty of threshold selecting in Canny algorithm, an improved method based on Otsu algorithm is proposed to choose the threshold adaptively and simultaneously. Firstly, guided by the gradient histogram of the test image, all the pixels are divided into three classes. Secondly, based on the improved Otsu algorithm, an evaluation function is defined to describe the mean square error among the three classes. Finally, both the high and low thresholds are selected adaptively and independently by searching the maximum values of the evaluation function. Artificial parameter setting is not necessary in this method. Compared with the results from traditional Canny method and Direct Otsu method, the method shows great advantage in extracting the real edges from different images, especially low contrast ones.