Denoise nl_means script

[EmmaRenauld]

Putting this here so that I don't forget this work in progress. See issue #939

[pep8speaks]

Hello @EmmaRenauld, Thank you for updating !

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2024-10-31 12:33:04 UTC

[codecov]

Codecov Report

Attention: Patch coverage is 69.86301% with 22 lines in your changes missing coverage. Please review.

Project coverage is 68.79%. Comparing base (9d94226) to head (38415f3).
Report is 82 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #972      +/-   ##
==========================================
+ Coverage   68.73%   68.79%   +0.05%     
==========================================
  Files         429      433       +4     
  Lines       22262    22500     +238     
  Branches     3322     3049     -273     
==========================================
+ Hits        15301    15478     +177     
- Misses       5674     5717      +43     
- Partials     1287     1305      +18     

Components	Coverage Δ
Scripts	69.63% <64.91%> (-0.02%)	⬇️
Library	67.59% <87.50%> (+0.16%)	⬆️

[mdesco]

I tested all options and it works.

Since, all calls Dipy stuff, I feel ok with this. And we go back to what we had in the past with proper support of Rician noise and Piesno. Not perfect but can be quite useful in advance context.

Would love to make this PR go to the master quickly as it would be extremely useful for my IMN530 group.

[EmmaRenauld]

Thanks! I'll check your comments soon.

I was also wondering: I only used the parameters that were used in the bitbucket for piesno, but there are other parameters:

alpha=0.01, step=100, itermax=100, eps=1e-5

(See https://github.com/dipy/dipy/blob/master/dipy/denoise/noise_estimate.py)

Should I add those parameters too?

[mdesco]

Good question @EmmaRenauld... not sure we should give access to these advanced parameters. I would not know how to tweak them myself... I guess an expert that would wish to play with this would probably write his own python script and call dipy directly, no? @arnaudbore thoughts?

Ping me if you want to chat about it.

[arnaudbore]

LGTM, quick comments