“…It is quite evident now that wavelet subband coefficients of natural images have highly non-Gaussian statistics [14], represent heavy tailed distributions. The usefulness of heavy-tailed priors for natural images are already exercised in image denoising [13,15], motion deblurring [12], blind deconvolution [1,16], image restoration [17,18,19] and video matting [20]. Some example of this priors class includes the generalized Gaussian [21], the Jeffrey's noninformative prior [15], and the Gaussian mixture (GM) [17].…”