Speckle Reduction of SAR Images in the Bandlet Domain

Synthetic Aperture Radar (SAR) images are inherently affected by multiplicative speckle noise, which is due to the coherent nature of the scattering phenomenon. This paper deals with the speckle reduction using the bandlet transform combined with the adaptive sigmoid thresholding. The operation needs to provide multiscale transform. We use the Undecimated Discrete Wavelet Transform (UDWT) and apply the bandlet transform on each resulting scale. Numerical tests applied on Lena image contaminated with multiplicative noise show that our method provides improvement both in terms of image visual fidelity and in terms of Peak Signal-to-Noise Ratio (PSNR). Comparisons are made with the standard wavelet transform and the shift invariant discrete time wavelet transform.