On the basis of the combined method presented in Reference[7],an improvement was implemented by exploiting the characteristics of both wavelet thresholding denoising and spatial Wiener filtering.After BayesShrink thresholding denoising in wavelet domain to obtain a pre-denoised image,the noise variance was estimated for the following Lee filtering.The optimal noise variance estimation for Lee filter was given by numerical computation.Experiment results show the improvement of the proposed approach in terms of MSE,signal-to-noise ratio (SNR),as well as adaptability to different images.