Speckle Denoising With NL Filter and Stochastic Distances Under the Haar Wavelet Domain

Abstract

Synthetic aperture radar SAR imaging systems have a coherent processing that causes the appearance of the multiplicative speckle noise. This noise gives a granular appearance to the terrestrial surface scene impairing its interpretation. The similarity between patches approach is applied by the current state-of-the-art filters in remote sensing area. The goal of this manuscript is to present a method to transform the non-local means (NLM) algorithm capable to mitigate the noise. Singlelook speckle and the NLM under the Haar wavelet domain are considered in our research with intensity SAR images. To achieve our goal, we used the Exponential-Polynomial (EP) and Gamma distributions to describe the Haar coefficients. Also, stochastic distances based on these two mentioned distributions were formulated and embedded in the original NLM technique. Finally, we present analyses and comparisons of real scenarios to demonstrate the competitive performance of the proposed method with some recent filters of the literature.