To improve the accuracy of boundary locations and region homogeneity, a approach based on simultaneous autoregressive model and wavelet -transform is proposed in this paper. This technique contains three stages: feature extraction, pre -segmentation and post-segmentation. Firstly, texture features are extracted by using the simultaneous autoregressive model, then the original image is segmented initially using the k-means clustering algorithm on the pre-segmentation stage. The post-segmentation is performed based on the result of pre-segmentation, in order to extract original image texture feature based on wavelet-transform. then the minimum distance classifier is used to classify the uncertain pixels, so the segmented image is achieved. Compared with traditional method, the present approach shows visible improvements both in diminishing segmentation error and in increasing boundary precision and region harmony.