Contour let-BasedMethod for Speckle Reduction with Adaptive Estimation of Noise Level

Abstract

Synthetic aperture radar (SAR) and ultrasonic images are inherently affected by speckle noise, which is caused by the coherent nature of the scattering phenomena. This paper presents a contourlet-based method for speckle reduction with an adaptive method for noisethreshold level estimation in a homomorphic framework. The method starts with the generation of many random images simulating the standard deviation level of the logtransformed speckled image. Different contourlet threshold levels are then calculated based on such simulations. Different contourlet coefficients of speckled images are thresholded by their corresponding pre-calculated contourlet thresholds. An exponential operation on the reconstructed output after thresholding is used to simulate the final homomorphic antilogtransformation stage and to obtain the de-speckled images. Unlike other classical and recent de-speckling methods, the despekled images indicate clearly the superiority of the proposed method for speckle reduction, especially for SAR images which possess a lot of detailed textures.