Abstract: According to excellence and shortcoming of conventional median filter algorithm，an algorithm called adaptive weighted median filter algorithm based on similarity function is proposed.The method works in this way：Firstly，noise detection is employed to determine noise points in image，then the size of filtering window is adaptively adjusted according to the number of noise points，and the pixels are adaptively grouped in an ingenious way based on the similarity and weight value of every pixels is calculated adaptively.Finally，a new weighted median image smoothing algorithm is used to filter noise pixels.The simulation experiment shows that this algorithm both may effectively eliminate the noise，and may protect image detail well.Therefore it has better filtering performance than the conventional median filter algorithm.