Compounding segmentation method for SAR images

Fast and accurate segmentation of synthetic aperture radar (SAR) images is important to SAR automatic target recognition (ATR). In this paper, a new compounding segmentation method is presented, based on the two-parameter CFAR segmentation method and Markov Random Field (MRF) -based segmentation approach. Following the initial segmentation obtained by the two-parameter CFAR method, the MRF-based approach is exploited only on the edge points of the targets and shadows. A comparative experimental analysis, using MSTAR (Moving and Stationary Target Acquisition and Recognition) images, of the proposed method against the other two approaches showed that the compounding segmentation method can achieve faster and more accurate results, which can be of practical use in the process of SAR ATR.