Statistical edge detectors applied to SAR images

Abstract

Segmentation is a crucial step in automatic interpretation of images. However, the presence of speckle in Synthetic Aperture Radar (SAR) images makes their segmentation very difficult. Several methods of segmentation have been developed specifically to improve the interpretation of SAR images. In this paper, we present edge detectors applied to SAR images. First, we present the detector Ratio of Averages (ROA), which is based on the normalized ratio computation of intensity. But, this detector is optimum in the case of mono-edge. Second, we present the detector Ratio of Exponentially Weighted Averages (ROEWA), which is optimized for multi-edge model, in the Minimum Mean Square Error (MMSE) sense. Finally, we give experimental results of these operators on optical image, simulated SAR image and on ERS-1 SAR image