A comparative study on SAR images speckle reduction in a wavelet transform framework

Speckle is a troublesome noise, disturbing radar SAR image interpretation and target classification. For reduction of speckle, numerous speckle filters have been proposed. Both suppression of speckle in a uniform area and preservation of edges have been pursued in these filters. In these years, the wavelet transformation is widely applied to images analysis, where the multi-resolution analysis by Mallat 1989) gives the theoretical basis. In this communication, we mainly study and compare two types of filters in a DWT framework: the ones based on the "à trous" algorithm of Dutilleux (1987), and those we develop using the "Mallat" algorithm (1989). For the first type, we apply either the Gaussian Gamma MAP and the Laplacian Gamma MAP filters to the wavelet coefficients got from the 'à trous' algorithm and we obtain the filtered image by the filtered images of the wavelet coefficients. For the second type, we use the "Mallat" algorithm with different wavelet bases instead of the "à trous" algorithm and we develop and discuss the case of Haar, Daubechies 4, 6, 8 and Biorthogonal (linear B-Spline and Battle Lemarie) Wavelets in two dimensions and show how they can be used for speckle reduction of SAR images.