SAR image segmentation based on the two-dimensional Pseudo Wigner-Ville distribution

The multiplicative nature of the speckle noise in SAR images is a big problem in SAR image segmentation. A novel method of automatic segmentation of local SAR images based on the two-dimensional (2D) Pseudo Wigner-Ville distribution (PWVD) is proposed. Firstly, we put emphasis on the problem of expensive cost and the interference of cross terms to SAR image segmentation. Then, we propose local 2D-PWVD transform to SAR image. The resultant data can be arranged as N energy maps which have the same size as original image making use of 2D local energy maps of every pixel of original image. The different processing to the resultant data in frequency dimension is adopted according to different energy maps. Finally, the marker region image is formed by virtue of adding the filtered energy maps. The segmentation experiment for MSTAR chips are taken and the comparison between segmentation image of our method and segmentation image of min error, between segmentation image of our method and segmentation image of max energy algorithm have been made. The result shows that the segmentation image of our method obtained by the presented algorithm preserves the abundant detail information of source so as to demonstrate the validity of the proposed method.