A New Region Gaussian Background Model for Video Surveillance

Background subtraction has been widely used in many intelligent video surveillance systems. It uses a background model to achieve target recognition by removing the background pixels of video images. Because of some very strong noises, the method is often hard to work well. For this reason, this paper proposes a novel background subtraction based on regional Gaussian model. Firstly, a modified region labeling algorithm is proposed to compute the sizes of residual disturbed regions left by color Gaussian background model. Secondly, the model is built based on statistic sizes of disturbed regions. The experimental results show that the region Gaussian background model can be more effective and robust to the strong noises of background than traditional color Gaussian model.

[1]  Georgios Tziritas,et al.  Adaptive detection and localization of moving objects in image sequences , 1999, Signal Process. Image Commun..

[2]  Wei Zhi-qiang.  A Moving Object Detection Method Based on Self-Adaptive Updating of Background , 2005 .

[3]  Zhang Qi-heng.  Real-time target detection algorithm under complex open country scene , 2006 .

[4]  Ming Zhang,et al.  Multiple Dimension Chrominance Model for Background Subtraction , 2005, Computational Intelligence.

[5]  Joan Serra,et al.  Image segmentation , 2003, Proceedings 2003 International Conference on Image Processing (Cat. No.03CH37429).

[6]  Viktor Öwall,et al.  Background Segmentation Beyond RGB , 2006, ACCV.

[7]  Frank Y. Shih,et al.  Image Segmentation , 2007, Encyclopedia of Biometrics.

[8]  Larry S. Davis,et al.  Efficient non-parametric adaptive color modeling using fast Gauss transform , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.