An optimal edge detector for bridge target detection in SAR images

Abstract

This paper describes an optimal edge detection method for synthetic aperture radar (SAR) images, which combines the ratio and gradient of averages (RGoA) edge detector and the optimal threshold selection method. We call the method modified RGoA (MRGoA). In MRGoA, two edge strength maps, ratio edge strength map (RESM) and gradient edge strength map (GESM), are computed by using the RGoA method. RESM can be considered as a normalized image, and GESM is a gray-level image. The ratio and gradient thresholds can be optimally selected by using histogram based methods, such as maximum correlation criterion (MCC). Finally, edge maps are obtained by using the ratio and gradient thresholds. In the proposed method, the criterion for determining edge pixels is somewhat different from that in the RGoA edge detector. Experimental results on SAR images have shown the effectiveness of our proposed method.