Robustness of Spann-Wilson Segmentation on Sar Imagery

Abstract

Previous studies have demonstrated that the Spann- Wilson algorithm has desirable quantities in segmenting noisy and textured images. However, the Spann-Wilson algorithm was originally tested for use with additive, and not multi- plicative, noise. Few works have systematically addressed the quantitative performance of this algorithm with speckle. This paper briefly describes the performance of this algorithm on simulated SAR images with varying degrees of speckle (one- to four-looks and varying amounts of white noise). One hun- dred forty-eight test images were considered, of which the al- gorithm segmented most without any adjustment to the algo- rithm's parameters.