In millimeter wave image restoration,L-R（Lucy-Richardson）algorithm is a simple and effective nonlinear method.However,when the noise can not be neglected,it is difficult for L-R algorithm to get good restoration.As an improved self-snake model has a merit of maintaining image edge and features and can de-noise effectively for the millimeter wave image having fewer data and lower resolution,a novel restoration method is proposed based on the improved self-snake model and L-R algorithm.It first de-noises by employing the improved self-snake model which edge stopping function is made of local variance,and then restores images by using L-R algorithm.The modified algorithm reduces the influence of noise on L-R algorithm effectively by using de-noise algorithm based on the improved self-snake model.The imaging results of experiment data show that the modified algorithm proposed in the paper improves the L-R algorithm performances of the signal to noise ratio and relevance,and it can be used in millimeter wave image with much noises.