Abstract: The noise in MRI image is a multiplicative noise with non-zero mean and obeys Rician distribution. However, many existing MRI image denoising algorithms directly apply methods which are right for additive white Gauss noise with zero mean. They denoise sub-optimally. An optimal bias corrected MR images denoising approach is present based on non-local means filter in DCT sub-space. Firstly, the characteristics of Rician noise in image are analyzed and the method to estimate noise bias in the space of squared magnitude MRI image is discussed. Then, the proposed MRI image denoising algorithm is utilized to reduce the Rician noise. Validation work is carried out on simulated images and the results demonstrate that the proposed method can denoise effectively and it can retain anatomical structures of organs and tissues while remove Rician noise in image.