Digital image processing is widely used in all fields of life, along with the higher requirement of the image transmatting information, it has been further developed and pretreatment of the noise processing is the key section. Classical methods for image denoising include the average method and median filtering, Wiener filtering, but these algorithms in noise processing has brought a series of missing details. And there are so many methods that lead to spending a lot of time selection algorithm to the specific noise, but not necessarily to get the desired effect. In this paper, the effect of the three methods of mean value method, median method and morphological method is analyzed. And combined with simulation results to adjust the parameters can obtain better quality image. The final result, morphological processing noise has more universality and superiority.