Segmentation of SAR images

Based on an SAR speckle statistical model, the authors investigate the feasibility of segmenting SAR images based on a gray-level histogram for one-look and multilook processed SAR images. The classification errors are evaluated and a procedure for the multilevel thresholding of SAR images is devised. The misclassification probabilities for one-look and four-look processed images are calculated and presented, and an unsupervised segmentation algorithm is given. A simulated one-look intensity image and a four-look SIR-B amplitude image are used for illustration.