Sea-Land Coarse Segmentation with Two and Three-Term LRS Decompositions in Multisquint Spaceborne Sar Imagery.

From the multisquint spaceborne Synthetic Aperture Radar (SAR) images, a novel method with low computing cost is proposed to perform the sea-land coarse discrimination. Both the two and three-term Low Rank and Sparse (LRS) decompositions are used, where the sparse and error terms are profited, respectively. The differences between each image pairs are then stacked and averaged. A thresholding test is finally operated. The experiment based on the real TerraSAR-X data preliminarily reveals that the land and sea areas could be roughly separated. The application includes avoiding territorial false alarms when conducting maritime surveillance in the sea-land junction region.