A Multi Modal Content-Based Copy Detection Approach

With the rapid development of network and multimedia technology, more and more digital video emerged, followed by the problem of copyright protection of digital video. Content-based copy detection methods (CBCD) have developed into a promising technology video monitor and copyright protection method.In this paper, a new content-based copy detection framework is proposed that these two aspects of the video and audio improvement and innovation.1、For the eight complex video conversions, we proposed robust global characteristics of DCT and local features DCSIFT video feature combination, the two complement each other. The binary characteristics DCT, uses fast indexing which is local sensitive hash (Locality Sensitive Hash, LSH), and for the local feature DCSIFT, a classic BoW index (Bag of Words) is used.2、For the picture-in-picture video conversion, proposed an effectively PIP detection algorithm.3、For the seven complex audio conversions, the improved audio characteristics EDF (Energy Difference Feature) based on the MFCC uses inverted index plus hash algorithm.4、For the matches between the video segment, proposed improved temporal pyramid matching algorithm, and then use the new level fusion algorithm fusion separate result of the audio and video, to thereby obtain a final result.Experiments on TRECVID datasets show that the results generated by this proposed framework is much better than the TRECVID average results in2011, reached the leading level.