Mixed local-global criterion for image denoising in the wavelet domain
Francisco Javier Garcia-Ugalde, Bohumil Psenicka · IEEE International Conference on Signal and Image Processing · 2007
This paper presents an image denoising method based on a two-step empirical Bayes approach. A linear minimum mean squared error-like estimation is performed to estimate the wavelet coefficients of the denoised image. These coefficients rely on a suitable estimation of the variance of the wavelet coefficients for the clean image. The later uses maximum likelihood estimation over a local neighborhood. As opposed to the approach presented in [3], the estimation of the variance of the coefficients for the clean image is performed only at locations corresponding to father and descendant wavelet coefficients greater than a threshold T. Thus, the proposed method is based on a mixed local-global criterion in the wavelet domain and the results achieved are among the best reported in the literature.