Abstract: Based on the multiscale, multidirection and shift invariance features of Nonsubsampled Contourlet transform(NSCT), A new algorithm for image denoising based on NSCT is proposed.In this paper.first the image is decomposed in different scales and orientations using NSCT.Then the threshold is derived according to each layer decomposition coefficients, which is used in image denoising.Furthermore.the processed coefficients are inversely transformed to get the denoised image.Experimental results show that the proposed algorithm has the better ability of image sparse expression than that of other methods in wavelet transform(WT) or contourlet transform(CT).It can not only eliminate the pseudo-Gibbs phenomena and noise effectively.but also has good effect and higher values of peak signal-to-noise ratio(PSNR), it preserves the detail and texture of original image perfectly as well.