A motion detection technique utilizing the size of encoded frame

This paper deals with a method on video analytics. In case of investigation crime, investigators check videos of the security cameras (CCTVs: Closed-Circuit Tele Visions). It is especially important when there are no witnesses at a scene. Recently, the number of CCTVs for crime prevention has increased rapidly in Japan for the purpose of establishing safe society. Currently investigators manually check videos, therefore an increase of the number of the CCTVs leads to an increase of their workload. Under these situations, researches relating to the automatic analysis of videos are more and more conducted. However, on the automatic analysis, detection omissions could occur when videos are recorded under bad condition such as low illumination and a faraway object. The calculation cost is also computationally expensive. Then we studied a method which is less detection omissions and can be easily accomplished. We visualized the size of encoded frame with video compression technique using the inter-frame coding such as H.264/MPEG-4 AVC, and emphasized motion in a video. This method helps to check the scenes which includes the motions. In conclusion, using this approach to long videos including less motion can reduce the burden of the investigators who check videos.