Wavelet neural network(WNN) was used to detect the noise-pixels and distinguish them from noise-free pixels.Then the noise-pixels were further classified by corresponding threshold and assigned the coefficients. Finally

the median filter was combined with the coefficient for the output.The proposed method is able to effectively eliminate the impulse noise while preserving more fine details.Visual evaluation and detailed statistical analysis(PSNR and SNR) prove that the proposed method is more improved than the conventional filters.Compared with the traditional median filter and the adaptive-length median filter

the method increases PSNR by 11.8 dB and by 7.59 dB respectively for the image with 20% impulse noise.