Adaptive wavelet thresholding for image denoising and compression

The first part of this paper proposes an adaptive, data-driven threshold fo r image denoising via wavelet soft-thresholding, The threshold is derived i n a Bayesian framework, and the prior used on the wavelet coefficients is t he generalized Gaussian distribution (GGD) widely used in image processing applications, The proposed threshold is simple and closed-form, and it is a daptive to each subband because it depends on data-driven estimates of the parameters. Experimental results show that the proposed method, called Baye sShrink, is typically within 5% of the MSE of the best soft-thresholding be nchmark with the image assumed known, It also outperforms Donoho and Johnst one's SureShrink most of the time. The second part of the paper attempts to further validate recent claims tha t lossy compression can be used for denoising, The BayesShrink threshold ca n aid in the parameter selection of a coder designed with the intention of denoising, and thus achieving simultaneous denoising and compression. Speci fically, the zero-zone in the quantization step of compression is analogous to the threshold value in the thresholding function. The remaining coder d esign parameters are chosen based on a criterion derived from Rissanen's mi nimum description length (MDL) principle, Experiments show that this compre ssion method does indeed remove noise significantly, especially for large n oise power, However, it introduces quantization noise and should he used on ly if bitrate were an additional concern to denoising.