Automatic Thresholding for Edge Detection in Sar Imagery

A few edge detectors are derived from the contrast ratio edge detector to extract linear features from SAR imagery with a constant probability of false alarm. But all of these detectors need one or more thresholds, which are generally predefined or selected by experience, to determine that the pixel belongs to an edge or not. In this paper, we propose a method to automatically determine the threshold according to the image feature based on ratio edge detector instead of predefining it as before. It introduces the principle of optimal entropy into the edge detection of SAR image, which classifies the intermediate result image of ratio-based edge detector into two classes: edge and non-edge. The presented method performs simply and fast, needless of manual intervention, which can avoid the influence of human factors. * Corresponding author.