Watershed identification of polygonal patterns in noisy SAR images

This paper describes a new approach to pattern recognition in Synthetic Aperture Radar (SAR) images. A visual analysis of the images provided by NASA’s Magellan mission to Venus has revealed a number of zones showing polygonal-shaped faults on the surface of the planet. The goal of this paper is to provide a method to automate the identification of such zones. The high level of noise in SAR images and its multiplicative nature make automated image analysis difficult and conventional edge detectors like those based on gradient images, inefficient. We present a scheme based on an improved watershed algorithm and a two-scale analysis. The method described extracts potential edges in the SAR image, analyzes the patterns obtained, and decides whether or not the image contains a “polygon area.” This scheme can also be applied to other SAR or visual images, for instance in observation of Mars and Jupiter’s satellite Europa. EDICS category: 2-SEGM (Segmentation)