Topology of high-contrast patches in SAR images

Abstract

With continuing growth in sensor capabilities and database complexity, the analysis and understanding of large data sets becomes increasingly important. Numerous researchers have recently explored the geometry and topology of high contrast patches in natural images in an attempt to understand the underlying manifold of the data. The results show that the majority of natural image patches are best represented by corners and edges, as one would expect from visual inspection. In this paper, we extend this analysis to log-magnitude SAR images from the MSTAR database. Our results show that the most representative high contrast patches in SAR images lie among the clutter however methods extracting target patches only show results more similar to that obtained for natural imagery. Contrary to their natural image counterparts however, high contrast patches in SAR imagery lack a significant geometric structure.