The SURE Approach to SAR Image Denoising Based on Multiscale Bandelet Contextual Model

Abstract

In this paper, a new denoising method is presented for Synthetic Aperture Radar (SAR) image, based on Stein's unbiased risk estimate (SURE) and on multiscale orthogonal bandelet domain. Unlike most existing denoising algorithms, A key point of our approach is that, using contextual model to compute contextual values of bandelet coefficients and then computing SURE thresholding according to these values. Multiscale orthogonal bandelet, a multiresolution geometry analysis tool, uses an adaptive segmentation and a local geometric flow suited to capture the anisotropic regularity of edge structures and provide an optimal representation of noisy SAR image. SURE threshold is used to handle outliers and heavy-tail noise, and it aims to minimize the mean-squared error between the true and restored image. Experimental results using real SAR image demonstrate that the approach can remove the speckle noise efficiently and preserve edge of SAR image better.