In order to effectively eliminate salt & pepper noise, Gaussian noise and even mixed noise in a noise image, a new image denoising algorithm was put forward based on the advantages of Wiener filtering and the features of all components of wavelet decomposition. This algorithm firstly conducted wavelet transform and separated 1 low frequency component and 3 medium & high frequency components, and then carried out self-adaptive wiener filtering for the low frequency component. It then extracted edges of the 3 medium & high frequency components by Canny operator, and finally reconstructed the 4 processed components and formed the de-noised image. Simulation results showed that the proposed algorithm had better performance in denoising common noise introduced by the scanner, with PSNR value more than 20 dB. Especially for Gaussian noise and mixed noise, it got relatively better PSNR value, 1~8 dB higher than Wiener filtering, soft threshold wavelet filtering and [9] algorithm. Combined with the wavelet transform and the Wiener filtering, this outstanding image denoising algorithm can better denoise various kinds of noise in noise images.