Hi Folks
I've been following this discussion.
Regarding the ringing phenomenon, I remember some work from Belgium: http://www.visielab.ua.ac.be/publications/correction-gibbs-ringing-diffusion-mri-data-using-total-variation-regularization which discussed the mechanism (and proposed a solution).
With regard to positive definite constraints: I just wanted to pick up on Rob's point..... "it has been shown that positive definite fitting does not have much if any statistical advantage over fitting with well determined weights" . Do you have a reference for that please?
Thanks,
Derek
|