Nonlocal Model-Free Denoising Algorithm for Single- and Multichannel SAR Data

Among the large number of synthetic aperture radar (SAR) image despeckling approaches existing in literature, nonlocal (NL) filters have received a desirable boost. However, often NL define similarity criterion based on model assumptions, such as fully developed speckle model. This assumption may not be verified high-resolution images urban environments. To address this issue, standalone model-free framework is proposed article. The presented approach provides generic for denoising variety SAR products, from single-intensity/amplitude to polarimetric and interferometric data. In particular, method empirical distributional between patch containing pixel recovered similar candidate pixel. decide whether patches follow distribution, Kolmogorov–Smirnov test adapted. Finally, restoration process aggregates selected pixels their relative importance derived distribution similarities. mitigate blurring effect preserve resolution, inhomogeneity ratio used perform bias reduction step. designed filter was tested different products results show that proves an unbiased able structures textures. It works automatically efficiently with single multilook (and multichannel) images.