Background-Removal plugin for ImageJ and ImageJ2 (Fiji)

ImageJ is a fully java-based software for image processing in fields such as physics and biology. Our plugin aims to remove the background from a stack of images using the Low Rank + Sparse + Noise method from the article Greedy Bilateral Sketch, Completion & Smoothing and inspired by the Python code of the GreGoDec algorithm. The original Matlab code from Tianyi Zhou has been forked by Andrews Sobral here.

• Low Rank => Background matrix
• Sparse => Moving objects martrix
• Noise => Disruption and errors matrix

Example of Low Rank + Sparse + Noise method

Original images:	Background images:	Sparse images:	Noise:

Installation tutorial

Requirement: ImageJ (Fiji) / Apache Maven and Java

Directly installing the .jar file as plugin in ImageJ

You can get the .jar of this plugin from the plugins folder here.

From the dedicated Fiji update site

You can add our update site [here](https://sites.imagej.net/FattaccioliLab/) to the Fiji Updater, following this tutorial : https://imagej.net/update-sites/following

Using Maven

From the project folder with Maven :

cd  Background-Removal-Plugin---ImageJ

mvn clean package -Denforcer.skip=true

cd  target

The file is named BackgroundRemoval-0.1.0-SNAPSHOT.jar in the project folder and GreGoDec-v0.1.jar in the ImageJ website. After you get the .jar plugin, you just need to paste it into plugins Fiji.app folder.

How to use it ?

You should now see Background Removal in the Fiji Plugins bar:

After clicking, the GUI appears, you can select a file with the associated button (opens the same files as Fiji) and swipe to see all frames:

• The gray clock changes color depending on the estimated time needed to Finalize the calculation, approximately for default parameters:
  • 🟢 Green clock < 5 seconds.
  • 🟠 Orange clock < 20 seconds.
  • 🔴 Red clock >= 20 seconds.

Parameters like Rank, Power, Error tolerance, Thresholding mode, Tau and k (numbers of greatest singular values in the SVD calculation) can be chosen to influence the output. You can choose Preview to see the result for the first 100 frames (or less if the stack has less frames):

If the output preview is satisfactory, you can choose to Finalize and save it from the default Fiji GUI:

Credits

v0.1 : 2022 - 2023 : Lyes LAÏMOUCHE, Alina NOVIKOVA, Daniel SIMA (Master in Software Science and Technology (STL) at Sorbonne Université, First year project)