IMBS-MT
Parallel Multi-modal Background Subtraction

Domenico Daniele Bloisi


IMBS-MT (multi-tread) is an extended version of the IMBS (Independent Multimodal Background Subtraction) library that use parallel computation.

IMBS-MT is written in C++ and is designed for performing an accurate foreground extraction in real-time for full HD images.

IMBS-MT can deal with illumination changes, camera jitter, movements of small background elements, and changes in the background geometry.
Additional details can be found in the paper
D.D. Bloisi, A. Pennisi, L. Iocchi, "Parallel multi-modal background modeling", Pattern Recognition Letters, 2016. draft

IMBS-MT functional architecture.

Get IMBS-MT

IMBS-MT is provided under the LGPL license and without any warranty about its usability. It is for educational purposes and should be regarded as such. By downloading it you accept the above listed conditions.

IMBS-MT (compatible with OpenCV 3)

Source code can be found here


Prerequisites for IMBS-MT


Build IMBS-MT

IMBS-MT is provided with a CMakeLists.txt file and can be compiled by using CMake.

Linux

  1. Unzip imbs-mt.zip in a folder, e.g., /home/imbs-mt
  2. $ cd <imbs-mt folder>
  3. $ mkdir build
  4. $ cd build
  5. $ cmake ..
  6. $ make

Run IMBS-MT

IMBS-MT is provided with an usage example (main.cpp)

Linux

For video files

$./imbs-mt -vid video1.avi

For an image sequence (fps = 25 default value)

$./imbs-mt -img images/1.png

or you can specify the fps value

$./imbs-mt -img images/1.png -fps 7


Results

Background Modeling

IMBS-MT has been tested on 14 different image sequences from the Scene Background Initialization (SBI) data set.

The results obtained by IMBS-MT are shown below.

Foreground Extraction

Examples of foreground extraction using IMBS-MT Parallel Multi-modal Background Subtraction.
Darker gray pixels are shadow points. No morphological operations have been applied.

         
         
         
         
    

Raw data: Scene Background Initialization (SBI) data set

Raw data: Town Centre Dataset

Person Detection

Person detection using IMBS-MT Parallel Multi-modal Background Subtraction.
Red boxes are validated using HOG descriptors + SVM (functions provided by OpenCV).

Raw data: Town Centre Dataset


References

[1] D.D. Bloisi, A. Pennisi, L. Iocchi, "Parallel multi-modal background modeling", In Pattern Recognition Letters, 2016. draft

[2] D.D. Bloisi, A. Pennisi, L. Iocchi, "Background modeling in the maritime domain", In Machine Vision and Applications, Springer Berlin Heidelberg, vol. 25, no. 5, pp. 1257-1269, 2014. draft

[3] D.D. Bloisi, "Background Modeling and Foreground Detection for Maritime Video Surveillance", Chapter in Handbook on Background Modeling and Foreground Detection for Video Surveillance: Traditional and Recent Approaches, Implementations, Benchmarking and Evaluation, Chapman and Hall/CRC, pp. 14-1-14-22, 2014. draft

[4] D.D. Bloisi, L. Iocchi, "Independent Multimodal Background Subtraction", In Proceedings of the Third International Conference on Computational Modeling of Objects Presented in Images: Fundamentals, Methods and Applications, Rome, Italy, pp. 39-44, 2012. draft


About

IMBS-MT has been written by Domenico Daniele Bloisi and Andrea Pennisi.

The authors wish to thank Fabio Previtali for helping in developing the code.