A novel image-denosing method based on the mixture gaussian model

A novel image-denoising method in orthogonal wavelet domain based on mixture gaussian model is presented. Dependent relationship between parent-child wavelet coefficients is considered by using the mixture gaussian functions to model the joint probability density (JPD) function of the parent-child wavelet coefficients; and its unknown parameters are estimated by the algorithm of EM(Expected Maximum). According to the JPD model, the shrinking estimator is given with the help of bi-variate Bayes rules. So, the denoised image can be reached. Some comparisons are made between the common wavelet-based method and the suggested method. The simulation shows the suggested method can get good standard performance and excellent visual effect, especially for low SNR noisy images.