AN IMPROVED BACKGROUND SUBTRACTION ALGORITHM AND CONCURRENT IMPLEMENTATIONS
Abstract
This paper proposes a concurrent implementation of a background subtraction algorithm suitable for monochromatic video sequences. In the training period, a set of frames is used to determine an estimate of the scene background, as well as background noise. In the test period, each new frame of the video sequence is compared to the background model, and foreground objects are obtained taking into account the background noise estimate. Shadows and highlights are also detected and removed. Concurrent programming tools and techniques, such as multithreading, are used to implement a concurrent solution for this problem, and combined with other optimization tools (such as destination word accumulation for morphological operators and the use of lookup tables for shadow detection/removal) to achieve the highest frame rate possible. Experimental results comprise a comparison of the performance including progressively each optimization tool, and they show that the fully optimized algorithm is much faster than a naive sequential implementation. Also, experiments performed on two-way processor and on dual-core processor computers indicate that the proposed algorithm is suited for the recent multi-core technology.
References
J. C. S. Jacques Jr. , C. R. Jung and S. R. Musse , A background subtraction model adapted to illumination changes, IEEE International Conference on Image Processing (IEEE Press) pp. 1817–1820. Google ScholarP. L. Rosin and T. Ellis , Image difference threshold strategies and shadow detection, British Machine Vision Conf. pp. 347–356. Google Scholar- IEEE Computer Vision and Pattern Recognition 2, 246 (1999). ISI, Google Scholar
- IEEE Transactions on Pattern Analysis and Machine Intelligence 27(5), 827 (2005). ISI, Google Scholar
- Image and Vision Computing 24(5), 473 (2006), DOI: 10.1016/j.imavis.2006.01.018. Crossref, ISI, Google Scholar
- Pattern Recognition Letters 27(7), 773 (2006), DOI: 10.1016/j.patrec.2005.11.005. Crossref, ISI, Google Scholar
- EURASIP Journal on Applied Signal Processing 2005(14), 2281 (2005). Crossref, Google Scholar
I. Haritaoglu , D. Harwood and L. S. Davis , W4: Who? when? where? what? a real time system for detecting and tracking people, International Conference on Automatic Face and Gesture Recognition pp. 222–227. Google Scholar- IEEE Transactions on Pattern Analysis and Machine Intelligence 22(8), 809 (2000), DOI: 10.1109/34.868683. Crossref, ISI, Google Scholar
- IEEE Transactions on Circuits and Systems for Video Technology 12(7), 577 (2002). Crossref, ISI, Google Scholar
- Pattern Recognition Letters 26(1), 5 (2005). ISI, Google Scholar
- Pattern Recognition 38(11), 1937 (2005), DOI: 10.1016/j.patcog.2005.02.006. Crossref, ISI, Google Scholar
- IEEE Computer Vision and Pattern Recognition 1, 1182 (2005). Google Scholar
- EURASIP Journal on Applied Signal Processing 2005(14), 2292 (2005). Crossref, Google Scholar
D. S. Bloomberg , Implementation efficiency of binary morphology, International Symposium for Mathematical Morphology pp. 209–218. Google Scholar- IEEE Transactions on Pattern Analysis and Machine Intelligence 25(7), 918 (2003), DOI: 10.1109/TPAMI.2003.1206520. Crossref, ISI, Google Scholar


