Summary: In order to solve the contradiction of reducing speckle and effective keeping details in speckle filtering for SAR (Synthetic Aperture Radar) images, a new image denoising method is presented, which is based on four-point interpolation subdivision wavelet. The method combines the wavelet with subdivision, and applies the rules of four-point interpolation subdivision to subdivision wavelet. In this paper, firstly the four-point interpolation subdivision wavelet is used to decompose the given SAR image. Then we reduce the noise of the wavelet coefficients by using adaptive threshold of Bayes estimation and the smooth threshold function. Finally, a technique is developed to reconstruct the denoising wavelet coefficients. The experimental results are evaluated by two evaluation indexes including equivalent number of looks (ENL) and edge save index (ESI). It is shown that the method proposed in this paper is more effective.