SAR Speckle Filtering Using Hybrid NonLocal Sigma Filter

Speckle noise in synthetic aperture radar (SAR) images reduces the accuracy of their interpretation. In this paper, a hybrid filtering technique based on the Sigma filter and the non-local mean filter is proposed. The neighborhood used for the estimation of the filtered pixels is selected by analogy to the non-local filtering principle. Non-similar pixels in the search area having high patch distances were discarded. The coefficient variation CV is employed to estimate the variability of the pixels used to select homogeneous pixels in the search area. Results showed that the proposed hybrid filter outperformed the Lee filter and the improved Sigma filter in terms of speckle reduction and spatial detail preservation. Airborne (AirSAR) and Space-borne (Sentlnel1) synthetic aperture radar SAR data were used for validation).