components


ImageJ Plugin for Non-Local-Means Filtering

kota's picture
Rating: 
4
Average: 4 (1 vote)
Package/Library: 
Author(s): 
Thorsten Wagner, Pascal Behnel
Description: 

A more modern approach for denoising / smoothening before segmentation, works like Gaussian blurring but preserves edges and boundaries. Listed in Fiji update sites.

Algorithm

Algorithm description is in this page [1].

Example usage

Localization of Membrane bound protein in Arabidopsis meristem was analyzed using the non-local-mean filter for refining its position [2].

impression

It's effect is somewhere between Gaussian blurring and anistropic diffusion.


References

Comments

1
david.rousseau's picture

Non Local Mean uses the self similarity of the object in the image to denoise. It is therefore very powerful in biology since objects are very self-similar. Think for instance to plant as stressed in reference below.

Rousseau, D., Chéné, Y., Belin, E., Semaan, G., Trigui, G., Boudehri, K., ... & Chapeau-Blondeau, F. (2015). Multiscale imaging of plants: current approaches and challenges. Plant methods, 11(1), 6.

Non local mean, in its basic version, screens for each single pixel in the whole image to seek for similar patches to help denoising. It is therefore much slower than any convolution viltering (Gaussian blur for instance).