Video Copy Detection Applying To IPTV

In recent years, as fast development of internet and Wi-Fi network, illegal video copies and piracy phenomenon are increasing quickly as well, which has done great damage to video owners'interests and internet video browsing and searching. Video copy detection, therefore, is becoming a new research focus. Content-based Copy Detection (CBCD), because of its many advantages, has been standing out and attracting more and more attention. On the other hand, Internet Protocol Television (IPTV), as a kind of very attractive internet application, has been acclaimed by users and is now under great promotion. This paper has proposed a concise but efficient method for video copy detection of IPTV applications. Our method take block gray histogram as basic video frame feature and find a fast but easy way to compute similarity between frames. This method can cover most video copy types in reality of IPTV except "Picture in Picture" problem. To conquer that, we find a special way to detect "Picture in Picture" cases making use of usual characteristics of IPTV. To improve detecting efficiency, frames are cut into shots using shot boundary detection methods, which also use the above-mentioned block gray histogram as basic feature to achieve good rate. Then shots are clustered, with a representative frame of one shot computed first, using the similarity between representative frames as shot similarity. Finally, shot clusters are indexed by K-means index tree, that will give different relative positions to clusters by their different similarities in the process of constructing the whole tree. So the search program can finish one search case with logarithmic time complexity. Our experiment results show that, the method has a satisfying searching speed, as well as good rate of covering and precision.