Denoise MRI images using sparse 3D transformation domain collaborative filtering

In the present work, a recent proposed transformation domain filter algorithm (BM3D) using the sparseness and self-similarity properties of the images is presented and adapted to reduce the specific Rician distribution noise in MRI magnitude images. The proposed R-BM3D method is compared to the state-of-the-art transformation domain filter method (ODCT3D), which is proved to be superior to the blockwise nonlocal means filter, the wavelet sub-band coefficient mixing method, and the anisotropic diffusion filter. Comparative experimental results indicate that the proposed R-BM3D is quite competitive in RMSE metric, filter time and visual inspection.