Study on wavelet-based image denoising

For the mixed Gaussian noise image, we proposed an approach of EM algorithm based on image histogram for noise variance estimation in this paper. Combined with wavelet analysis, this approach is used to process image denoising. After noise variance estimated, we process threshold on scale coefficients on the basis of the wavelet transform properties. And we make full use of the prior information of the noise and signal distribution in order to construct modified factors to modify wavelet coefficients. Experimental results show that this proposed method performs well on variance estimation and both the visual effect and SNR of denoised image are improved. A satisfying denoising effect is obtained.