Bayesian wavelet-based shrinkage for SAR images despeckling using generalized Gamma distribution
Pingping Huang, Heng-Chao Li, Pingzhi Fan · IEEE Asia-Pacific Conference on Synthetic Aperture Radar · 2011
In this paper, a non-homomorphic wavelet-based maximum a posteriori (MAP) despeckling of SAR images using two-sided generalized Gamma distribution (GΓD) is proposed to suppress speckle. Instead of the logarithmic transformation, the multiplicative speckle model is decomposed into an additive model with signal-dependent noise. And a locally adaptive MAP estimator of noise-free wavelet coefficient is derived by introducing a two-sided GΓD and a local Gaussian distribution as the statistically models for the wavelet coefficients of the additive signal-dependent noise and the reflectance image, respectively. Further, an approximated scheme is provided to estimate the parameters of the involving prior distribution. Finally, experimental results, carried out on the actual SAR image, demonstrate the validity of our proposed despeckling method.