关键词: 图像分割, 合成孔径雷达图像, 域马尔可夫模型, 非降采样Contourlet变换（NSCT）

Abstract: An algorithm of Markov Field（MRF） model based on Non-drop Sampling Contourlet Transform（NSCT） domain is proposed for the Synthetic Aperture Radar（SAR） image under complex background. Utilization of the advantages of both the MRF model and multi-resolution description of the information, Gaussian Model can be used in the characteristics field of each scale and Potts Model also used in the mark field of each scale. Then as the initial value, large-scale segmentation results can be directly projected to the small scale. In the experimental part, comparing with the classic threshold segmentation algorithm and Markov segmentation algorithm, the result shows that target can be accurately segmented while preserving more details.

Key words: image segmentation, Synthetic Aperture Radar（SAR） image, Markov Field（MRF） model, Non-drop Sampling Contourlet Transform（NSCT）