Comparison of Background Extraction Methods for Anomaly Detection

Background extraction is one of the pivotal tasks in automated video analysis. It forms the primary step in abnormal event detection in surveillance videos. In this paper, we present a comparative study of background extraction methods for application in traffic surveillance videos. Performance comparison is performed on real surveillance videos with different illumination and climatic conditions. Qualitative as well as quantitative evaluations on standard datasets show that adaptive mixture of Gaussians is the best among the compared algorithms.