Near-duplicate video clips detection with motion based video fingerprinting

摘要

Recent rapid development in video capture and broadband communication technologies have brought video content production into an explosive growth era. However, more and more near-duplicate video clips spread over the Internet. It not only lowers the retrieval efficiency but also wastes the network bandwidth and storage space. Therefore effective and efficient detection of near-duplicate video clips become a challenging issue. In this paper a video fingerprinting based scheme is proposed to detect the near-duplicate video clips. Video fingerprint consists of two motion features, including motion magnitude and motion activity. Experimental results on detecting near-duplicate video clips over several kinds of video data show the effectiveness and efficiency of the proposed combined scheme based on video fingerprint.