The speckle is omnipresent in synthetic aperture radar (SAR) images as an intrinsic characteristic. However, it is unwanted in certain applications. Therefore, intelligent filters for speckle reduction are of great importance. It has been demonstrated in several literatures that the non-local means filter can reduce noise while preserving details. This paper discusses non-local means filter for polarimetric SAR (PolSAR) speckle reduction. The impact of different similarity approaches, weight kernels, and parameters in the filter were analysed. A data-driven adaptive weight kernel was proposed. Combined with different similarity measures, it is compared with existing algorithms, using fully polarimetric TerraSAR-X data acquired during the commissioning phase. The proposed approach has overall the best performance in terms of speckle reduction, detail preservation, and polarimetric information preservation. This study suggests the high potential of using the developed non- local means filer for speckle reduction of PolSAR data acquired by the next generation SAR missions, e.g. TanDEM-L and TerraSAR-X NG.
Volumetric non-local-means based speckle reduction for optical coherence tomography