I'd say the transform as it should retain all the original data

Of course Transform domain. In fact, many of modern CAD based medical diagnosis systems are coming with inbuilt filtering that includes Fourier and other transforms...

DCT or Zernike moment based NLM filter provides better results....u can read New methods for MRI denoising based on sparsness by JV Manjon and P.coupe....

Dear Karamjeet,

 i had read alot of papers by Manjon and Coupe, they say that NLM means filter gives better results. Besides this, there is another approach which outperforms them, that is HOSVD. It gives better PSNR. Please find attached the relevant paper. 

Regards,

Dear Laraib Kanwal

The simple fact is there are lot of other approaches also....i give my reply about what u asked dear.....Another paper entitled, MRI noise estimation and nonlocal PCA by same authors represents very good results....the main advantage is that their approach is simple and easy to implement in c/c++ and in matlab. Moreover, in that particular paper which i have mentioned have relevant PSNR and MSSIM tables u can easily match ur results.

Thanks

Karamjeet SIngh

Thanks sir karamjeet, can you tell me how to load MRI image on matlab? I am downloading T1-weighted MRI image from brainweb. The image format is either .mnc or .rawb. Unable to open on matlab

download the manjon and p.coupe sir code and there is program that convert dcm files to tiff files....thanks