The image de-noising method via USFFT Curvelet transform is proposed in this paper. The method can preserve the detail information of high frequency of noisy image, and to avoid the phenomenon of edge slant during the image reconstruction. The USFFT Curvelet transform is used to the noisy image, the global threshold value is estimated in the Curvelet domain. The window technique is used to compute the shrinkage factor corresponding to pixel value. The de-noised image signal is obtained by using inverse USFFT Curvelet transform. The experimental results show that the algorithm obtains a higher PSNR value for gray and color images. It acquires a good de-noisy effect and visual effect, and preserves image edge well.