In this paper, a new algorithm based on three steps has been proposed for segmentation of SAR images. In the first segmentation step, the input SAR satellite image is proposed using CLA and the difference between the gray levels of pixels in each region of SAR image is reduced and a basic segmented image is produced which is used as an input image to the second step of the algorithm. In the second step which is an expletive segmentation step, a q threshold is selected as a gray level distance between the pixels of each region. If the gray level distance between two adjacent pixels is smaller than q, then these two pixels are related to same region and will take the same label. After labeling the regions, the average of the same labeled pixels of gray level is considered as gray level indicator of each region. Theresholding is the last stage of the proposed algorithm in segmentation process for identification of regions. In this step, all of the same regions, take the similar label. The proposed algorithm is evaluated on two types of SAR images: the simulated SAR images and the real SAR images. The results show that the proposed method has less error rate and higher accuracy in comparison with other methods for segmentation of SAR images.