Based on the statistical property of image noise and combining with BayesShrink theory, a method of image denoising based on wavelet domain Contourlet transform is presented. Using BayesShrink theory to estimate the threshold, considering the local correlation of the neighborhood, then improving the adaptive method of selecting threshold, finally obtaining the optimal threshold Ti,j [σX(LD)], this algorithm has implement the image denoising. Furthermore, analyzing the peak signal to noise ratio (PSNR) and its computational complexity. The simulation results show that the superiority of this algorithm which has obviously improved the visual effect and PSNR when compared to DWT- NABayesShrink method, DTCWT- BayesShrink method and CbATD method.