Abstract: A novel mixed iteration filtering algorithm is presented to reduce speckle noise in SAR image by combining the Lee filter with variable-size windows and self-snake diffusion, and an effect method to estimate noises variance of SAR image is given. Local statistical parameters of SAR image is first estimated, and then SAR image is smoothed by improved Lee filter with variable-size windows. Then self-snake diffusion is used to remove the remaining isolated points noise and edge noise in the image smoothed by Lee filter. The window size of Lee filter will be increased with the number of mixed iteration, and thus homogeneous regions of SAR image are smoothed better while image edge is protected. The experimental results show that, compared with several kinds of despeckling algorithms, the proposed algorithm has better performance in speckle suppression and edge protection.