Abstract: Curvelet is a kind of multi-scale transform, it provides a expression of stability, efficient and optimal, for target function with smooth-curve-singularity.A coefficients product image denoising method based on second generation curvelet is proposed, using coefficients of decomposition has the same characteristics in different scales, based on second generation curvelet.The result shows that the algorithm achieves better denoising than wavelet transform and simple thresholding methods.