A Video Hierarchy Analysis Framework Based on Motion
Yuxiang Xie · Journal of Chinese Computer Systems · 2007
Motion information is one of the most important features in video. A novel video hierarchy analysis framework based on motion is presented in this paper, which can be used to describe and retrieval video stream for reflecting the motion relation and camera motion information during the producing of the video. A new concept is proposed in this framework: Basic Motion Unit (BMU), which reflect the dominant camera motion type defined in MPEG-7. According to the above idea video can be segmented into the set of BMUs. Then the motion semantic schemes are discovered and mined based on BMU. Finally BMUs are cluster into Motion Activity Unit (MAU) with fast or slow motion rhythm according to the motion activity proposed in MPEG-7. In addition, the presented video analysis framework can not only capture the attention of cameraman but also provide a new approach for the efficient non-linear accessing of video streams.