Content-Based Video Copy Detection(CBVCD) is one of the most important technique for video copyright protection. The technique is extracting video summary without the visual features. And then, measuring the distance between the fingerprints to determine whether is a copy one.Based on analysing the video coding standards and the video visual characteristics, a new real-time video copy detection system based on compressed-domain fingerprint extraction and two-stage retrieve algorithms is proposed. Firstly, we extact the DC coefficients of the OM features combining with the motion vector. Then retrieving the matching levels based cluster framework to accelerate matching process.Experiments show that our algorithm is superior to some recent proposed algorithms both in robust and efficiency. So our algorithm has a higher theoretical significance and practical value. |