The soft threshold function will produce a constant deviation cause and the hard threshold function is not continuous cause image edges blur. In this paper a new semisoft threshold function method is proposed. This method processes the image by distinguish strong and weak edge. What’s more, the neighboring wavelet coefficients and Bayesian estimation method were incorporated into the estimation of the weak edge. Simulation results showed that the proposed algorithm had better visual effect and PSNR and EPI performance than many exiting thresholding methods. This method can protect the edge of the image to achieve a good denoising effect.