Processing of SAR data exploiting spatial, polarimetric and topological information

The Markov random field (MRF) approach is very useful in synthetic aperture radar (SAR) image classification and segmentation. Unfortunately, it exhibits the drawback that fine structures tend to disappear during the segmentation process due to the morphology of the cliques used for the region label model. In this paper, a new method is described that tailors the shape of the cliques to topological information. The validity of the approach is shown by the results obtained by applying the method to multi-polarimetric SAR images.