Application of Non-local Mean Filtering in Real Head MR Image

MR images have high resolution for soft tissues, which have been widely used for brain research. During the acquisition of MR images due to the interference of the hardware device and external environment, it introduces a amount of noises, which will affect the clinical accuracy of the diagnosis of the disease. The basic principle of non-local mean filtering is similar to the mean filter, but the non-local mean filter adds the weight value of each point to the calculation, so it can be ensured that when neighboring and very different points are averaged in the box, which also retains a lot of details on the edge of the image, so that the image will look more clear. In this paper, the non-local mean denoising algorithm is compared with three kinds of preserving and denoising algorithms: median filtering, anisotropic diffusion filtering, and bilateral filtering. It is verified that it has good performance in denoising while maintaining the edge, and applied in the denoising of real head MR images.