Abstract: A wavelet-based adaptive filtering algorithm for mixed noise is proposed.Firstly,median filter is used in image preprocessing to remove impulse noise.Secondly,the image is conducted two-dimensional wavelet decomposition,obtaining high-frequency and low-frequency sub image.Thirdly,according to the noise distribution of high-frequency sub image,designing new morphological filtering structure elements respectively,then a new threshold discriminant function is defined,adjust parameters is set up separately for high-frequency and lowfrequency sub image,so as to filter out residual noise.Finally,wavelet coefficients were reconstructed.The simulation results show that the algorithm is obviously superior to other filtering methods.