Real-time motion detection in H.264 compressed domain for surveillance application

Abstract

Camera motion detection is a useful functionality in video surveillance and many other areas, but with its widespread application in many areas, the compaction power has increased and needs to be addressed how to reduce it. So far, there has been a lot of research on pixel domain, but it is hard to operate with real-time and consumes very much computing power. However, since the compressed domain of block-based codec contains useful information for video analysis, this allows accurate video analysis at a low cost. The proposed method shows how to perform a motion detection accurately while significantly reducing the amount of computation compared to the pixel domain by using the DCT coefficients, motion vectors and other features in the compressed domain of the H.264 video. The proposed algorithm is designed for motion detection in video surveillance servers in the form of a cloud. The results of the experiment were shown to support validity.