Abstract: In this paper, an algorithm based on Local Gaussian Mixture Model of Dual-Tree Complex Wavelet Transform is proposed for image denoising. A classification is performed based on posterior probability of the Gaussian Mixture Model, with different class matching different Gaussian model. Then subband coefficients are estimated by using Maximum a Posteriori estimator. As approximate shift invariance and good directionality of DT CWT, it can remove “ringing" of the edge region. Experimental results show an improved denoising performance of PSNR and human vision in comparison with other methods.