Gradient-based hyperpixel segmentation for polarimetric SAR images

Superpixel segmentation has been widely used in the field of image segmentation due to its excellent performance. The simple linear iterative clustering (SLIC) method shows excellent performance in optical images and has been widely used in polarimetric SAR images. However, the initialization step of the SLIC method cannot accurately locate the class center, requiring multiple iterations to correct the errors. The improved watershed method (SCoW) is a simple and efficient segmentation method based on gradient thresholding discrimination, but it cannot be directly used in polarizing SAR images. Inspired by SCoW, in this paper, we propose a segmentation method which serves for the SLIC preprocessing step. First the edge information is detected by a CFAR edge detector, and then the edge information is used to initiate the segmentation. Experiments using real PolSAR data show that this method can reduce the number of iterations. The segmentation result is more consistent with the image boundaries in comparison with other three superpixel segmentation algorithms.