Study Of SAR Edge Detection Algoirthms

Edge detection is a fundamental tool in image processing and computer vision,particularly in the areas of feature detection and feature extraction. The gradient-basededge detector does not work well for SAR images for the existence of the multiplicativenoise in SAR images. This paper presents a constant-false-alarm-rate (CFAR) edgedetector based on ratio to extract thin edges applied to SAR images. False edge pixelstend to appear around true edges when we use the traditional rectangle bi-windowconfiguration. In order to avoid the poor smoothing effect of rectangle bi-window, wepropose an edge detector using Gauss-Gamma-shaped window function. The newwindow function can reduce false edge pixels around true edges, which is verified bythe analysis of the effective false maximum in the edge strength maps (ESM). We usenon-maxima suppression and hysteresis thresholding, which are in common use in edgedetection of optical images, on the ratio-based ESM to extract thin edges for SARimages. The receiver operating characteristic (ROC) curve is an efficient evaluator foredge detectors in SAR images. Synthetic SAR images with ground truth and a real SARimage are used to evaluate the detector using rectangle bi-window and the one usingGauss-Gamma-shaped bi-window on the criteria of ROC curves and visual effects. Theexperimental results prove the proposed detector is superior to the one using rectanglebi-window on the above criteria.