SPECKLE SUPPRESSION OF SAR IMAGES BASED ON WAVELET SINGULARITY DETECTION

摘要

A new algorithm for speckle suppression in synthetic aperture radar (SAR) images based on wavelet singularity detection is presented. The wavelet; coefficients of the SAR images that correspond to the regular parts of the signal are first selected by Mallat's wavelet transform modulus maxima (WTMM) approach. Then, the singularity of the residuary wavelet coefficients is detected in every scale by using the Lipsclutz-Holder exponents, which represents the singularity strength of noise. Speckles can then be suppressed by a weighted averaging lilter with coefficients determined by the Lipschitz-Holder exponents. The final image is formed by the fusion of the selected thematic signals and the Lipschitz-Holder exponent based weighted averaging parts. As demonstrated by the simulations, this approach can improve the denoising effect, in terms of speckle suppression, edge preservation, equivalent number of look (ENL), and radiometric resolution, as well as the visually-natural images generated.