Abstract: A novel segmentation method is proposed for MSTAR SAR target chips. First，feature extraction is implemented to acquire the attributed scattering center(ASC) feature vectors based on scattering characteristics. Then， a Markov random field(MRF) with the attributed scattering center feature is used to describ the spatial contextual information of different segments in the SAR image ofMSTAR SAR chips. The final segmentation result is obtained by combining an optimized algorithm oflabel costs. Experimental results validate the effectiveness of this improved algorithm.