Effect of H.264/AVC compression on object detection for video surveillance

Nowadays, video surveillance systems apply video compression to reduce bandwidth and storage cost. However, generally, these video sequences are the input, after decoding, for video analysis modules. H.264/AVC is the newest video standard and is assumed to be omnipresent in video surveillance systems in the near future. Since the video compression introduces artefacts, which influence the performance of these analysis modules, it is important to make a quantitative evaluation of this effect. Hence, in this paper we present the first quantitative analysis of the effect that H.264/AVC compression has upon a generally accepted moving object detection technique. We analyze different encoding schemes and show the influence on the object detection results for different representative sequences.