Fully Polarimetric SAR Image Filtering by Combining Block Matching and Lee Filter

Speckle filtering is a key preprocessing approach for the applications of polarimetric synthetic aperture radar (PolSAR) data. Most current PolSAR filtering algorithms can hardly get good balance in suppressing speckle and retaining image details. In this article, an algorithm for fully PolSAR image filtering is proposed by embedding the idea of the advanced block matching theory into the classical Lee filter, which ensures that image details can be well preserved after the significant reduction of speckle. First, taking each image block as the reference, a block matching step is taken to group those similar blocks in the nonlocal area; the Lee filter is employed to simultaneously process the center pixel of each block in the group, followed by an aggregation step which estimates the same pixel been grouped in different groups; then, the block matching step is taken again to regroup similar blocks by using the information of both the original image and the image filtered in last stage; again, the Lee filter is collaborated with the aggregation step in filtering the image. Despeckling experiments demonstrate that the proposed PolSAR filter can not only significantly suppress the image noise but also well preserve edges and textures.