Segmentation of Polarimetric SAR images using graph partitioning active contours

Abstract

In this paper, a novel segmentation method for Polarimetric Synthetic Aperture Radar (PolSAR) images is proposed. Image segmentation is formulated as a graph partitioning problem which is addressed by curve evolution. Graph partitioning active contours (GPAC) [1], which is flexible and can be classified as a region-based active contour framework, is applied to PolSAR images. In order to apply GPAC to PolSAR image segmentation, the pairwise dissimilarity between two arbitrary pixels in the image should be measured appropriately. The selection of dissimilarity measures is discussed. Synthetic and real PolSAR images are both used in the experiments to verify the proposed approach. A quantitative evaluation and comparisons of segmentation results, including iteration times of curve evolution, segmentation accuracies in term of pixel and object count, are provided. The proposed scheme has shown to be promising for PolSAR image segmentation.