Substantial noises have been added in medical images during its acquisition which adversely affects its quality and hence may leads to misdiagnosis. Therefore it is necessary to process these images in order to reduce noises before their interpretation and analysis. In this paper, we proposed a Contourlet based adaptive threshold method to reduce the noise of MRI images. We have compared the proposed method with other conventional de noising methods, and found that our method outperforms in terms of PSNR. Visual quality of images after denoising by using our proposed method is also superior as compare to other methods.