NLmeans unclear! - Githubissues

scilus / scilpy

The Sherbrooke Connectivity Imaging Lab (SCIL) Python dMRI processing toolbox

Other

54 stars 59 forks source link

NLmeans unclear! #939

Open EmmaRenauld opened 4 months ago

EmmaRenauld commented 4 months ago

Hey!

The mandatory argument number_coils in scil_denoising_nlmeans.py is never used!

We verify if it is >0 to decide between Rician and Gaussian. But you can put 1, 2, 4, it does change anything!
The call to dipy's estimate_sigma uses N=0, where N is the number of coils, so we always enforce 0. https://github.com/scilus/scilpy/blob/master/scripts/scil_denoising_nlmeans.py#L63

Who can help me decide what to do.

Other questions:

we take the median of the std, but it is not explained why not the mean.
All the comments in that script seem old and deprecated. There has been no change to this script since Neil Goyette's work in 2017 (see bitbucket: https://bitbucket.org/sciludes/scilpy/annotate/master/scripts/scil_run_nlmeans.py?at=master). Ex: It talks about bval but this script works on any data. Can I delete all the comments?