关键词: 合成孔径雷达（SAR）, 马尔可夫随机场（MRF）, Gaussian-Hermite矩, 最大后验概率（MAP）

Abstract: Markov Random Field（MRF） has lots of applications in SAR image segmentation.Because of the multiplication nature of speckle noise in SAR image，it is hard to get accurate segmentation，in this paper，a new fusion algorithm for unsupervised SAR image segmentation based on Markov Random Field（MRF） with Gaussian-Hermite Moments（GHM） is proposed.It gets initial segmentation with different orders of Gaussian-Hermite moments as SAR image features.Then，the first result can be as initial MRF model，by introducing an energy function-based weighting parameter between the two components，the accurate result is achieved by Maximum A Poteriori（MAP）.The proposed method applied in synthetic image and SAR image segmentation，and experimental results show that it performs well in miss-classification ratio，restraining the speckle noise，and performation in vision.

Key words: Synthetic Aperture Radar（SAR）, Markov Random Field（MRF）, Gaussian-Hermite Moments（GHM）, Maximum A Poteriori（MAP）