Page history Edit this page How do I edit this website?

Non Local Means Denoise


This is an ImageJ plugin for denosing images via the non-local-means algorithm described in


including the changes proposed by


It is numerically optimized and multithreaded. It works with all image types (RGB, 32 Bit, 16 Bit, 8 Bit).

This plugin was developed within the scope of a study work of Pascal Behnel and is maintained by Thorsten Wagner (Both are members of the Biomedical Imaging Group).


Sigma: The sigma of the noise.

Smoothing factor: In most cases, the default value (1) should not be changed. However, sometimes a more smoothed image is desired which can be achieved by selecting values > 1.

Auto estimate sigma: If this option is selected, the sigma is automatically estimated by the method descriped in Immerkaer, J., 1996. Fast noise variance estimation. Computer Vision and Image Understanding.

Both, the value for the (estimated) sigma and the smoothing factor are saved in the imagej preferences and are accessible with the keys “nlmean.sigma” / “nlmeans.smoothingfactor”.


The left image shows a noisy (sigma=25) image, and right image the denoised version using the non local means plugin with auto-estimated sigma and smoothing factor of 1:

clown-noise-25 clown-denoised


You could simply use our update site “Biomedgroup” to install the non local means plugin, or copy the jar file into your plugins folder.

How to cite

We think the best way is to cite the formal method and the used implementation: