Diffusion weighted magnetic resonance image (DWI) should be denoised effectively for the corresponding procedure due to its property of imaging and application.Different from the normal gray level image

the noise in DWI is distributed under the Rician distribution.The commonly used local denoising method is lack of the synthetic implementation of the statistical information of noise

especially the Rician noise in the DWI.This paper proposed a modified LMMSE restoration method used for DWI.The proposed method used the local information to estimate the parameter of the Rician noise and modified the LMMSE using the principle of the anisotropic filter.The simulation and experiment of the synthetic DWI and real human brain DWI dataset demonstrated that the proposed method can effectively remove the Rician noise compared to the commonly used local denoising method and improve the robustness and validity of the DTI.