The invention discloses a multi-tensor-based magnetic resonance diffusion weighted image structure adaptive smoothing method, relates to a magnetic resonance diffusion weighted image smoothing method, belongs to the field of medical image processing, and solves the problem that the accuracy of the fiber structure information of each obtained voxel is low because of poor noise suppression in the conventional method. The multi-tensor-based magnetic resonance diffusion weighted image structure adaptive smoothing method comprises the following steps: firstly, selecting related parameters, and setting an initial neighborhood radius; secondly, calculating the initial fiber structure information of each voxel; thirdly, calculating the weights of all voxels in the neighborhood radius on the voxel according to the fiber structure information, performing weighted smoothing on a magnetic resonance diffusion weighted image, and recalculating the fiber structure information of each voxel after the magnetic resonance diffusion weighted image is smoothed; fourthly, judging whether a stopping criterion for iteration is met or not, if not, expanding the neighborhood radius and continuing to performing the third step, or else, ending the calculating. The multi-tensor-based magnetic resonance diffusion weighted image structure adaptive smoothing method is applicable to processing the information of the magnetic resonance diffusion weighted image.