An Efficient Histogram-Based Method for Background Modeling

An efficient background modeling technique is a key step in surveillance or video content processing systems such as event detection, human action recognition and semantic indexing of video. In this paper, a simple but efficient modeling scheme, which is able to deal with static and dynamic backgrounds, is presented. First, each pixel is averaged by a group of dominant bins based on histogram analysis, then the background can be successfully detected and further classified the background into static or dynamic according to the statistic properties. We simulated for various challenge test videos, the experimental results shown that the proposed method is robust and very computational efficiency.