A GΓD based level set method for target segmentation in high resolution SAR images

Abstract

A level set method is proposed in this paper for target segmentation in high resolution synthetic aperture radar (SAR) images. The generalized gamma distribution (GΓ D) is a very flexible distribution which can model the diversity of a large range of scenes in high resolution SAR images. In the proposed target segmentation method, the energy functional is established based on the GΓD and the curve evolution is then implemented by the variational level set method. Experiments with MSTAR data show the proposed method can effectively extract target from high resolution SAR images. Compared with the k-means algorithm and the traditional level set method, the proposed method could obtain more accurate target segmentation results with a lower false alarm probability.