Abstract: Firstly, pseudo median filtering algorithm was improved. Factors of gray value and geometric distance of pixels are integrated into noise detection process so as to separate noise from image pixels. The noise is filtered by weighted filtering method. Then obtaining method of prior information of an improved non-local means filtering algorithm was improved. The noise image is conducted lifting wavelet transform. The low-frequency coefficients are selected by a new threshold function, and the selected low-frequency coefficients which are greater than the threshold are reconstructed to obtain a reference image. Weighted value of the reference image is computed and regarded as prior information. Finally, a filtering method for infrared image based on the two improved algorithms was proposed, namely using the two improved algorithms to filter noise in infrared image successively, then correcting the filtered image by the reference image. The experimental results show that the method has good filter effect for infrared image with high density noises.