For the existence of speckle in SAR image, it is very difficult to obtain good segmentation result with traditional methods. In the paper, a new method of SAR image segmentation is proposed with the combination of the features of SAR image and multi-resolution analysis of wavelet. First, texture features are abstracted in wavelet domain. Second, SAR image is filtered according to the feature of its distribution in the wavelet domain. At last, with Fuzzy C-Means clustering technique (FCM), SAR image is segmented according to the vector composed by wavelet texture features and gray level of SAR image filtered. The experiment results show that the new method is efficient in SAR image segmentation.