A New Canny-Based Edge Detector for SAR Image

Canny has proposed three continuous criteria to compare the performance of different filters: good detection, good localization and low-responses. Based on these criteria, the optimal filter, derivative of Gaussian function, for edge detection is selected. Unfortunately, it does not wok well when it was employed to detect edges of SAR image for the reason that the SAR image is polluted by the multiplicative noise which satisfied the Gamma distribution. In this paper, a new Canny-based edge detector for SAR image is proposed. The kernel function of edge detector is modified according to the noise, also preprocess is performed. Finally, the simulated results prove its superior to traditional Canny operator and others.