Spatially adaptive denoising based on mixture modeling and interscale dependencies of wavelet coefficients

In this paper, we propose a mixture modeling of wavelet coefficients for image denoising. A binary mask value is constructed using the parent-child relationship of wavelet domain. Using probabilities based on the significance map, probability weighted Wiener filter is proposed, and also we develop the method of selecting windows of different sizes around the coefficient. Experimental results show that our method outperforms other critically sampled wavelet denoising schemes.