A novel inverted index file based searching strategy for video copy detection

摘要: The demand of video copy detection system is growing rapidly, as the development online uploading and sharing. In past two decades, research concentrates on information extraction or feature building rather than fast searching strategy. However, number videos growing, in systems has become a big issue. this paper, we propose novel strategy for Inverted Index File (IIF) based by using fingerprinting technology. proposed approach consists parts — fingerprint matching fragment matching. To speed up process, table lookup operation utilized that rely counting matched sub-fingerprints instead Hamming Distance metric. For matching, all candidate used to more one with different similarity score. tested experimental content-based methods, distortion types database scale. Experimental results show achieve high accuracy, around 10 times faster compared conventional IIF method. Moreover, upscaling, rate methods further make it potential be large-scale systems.