An efficient content based video copy detection using the sample based hierarchical adaptive k-means clustering

摘要

Content-based video copy detection (CBCD) is very important for video copyright protection in view of the growing popularity of video sharing websites, which deals with not only whether a copy occurs in a query video stream but also where the copy is located and where the copy is originated from. In this paper, we present a video copy detection scheme based on local features which can deal with very large databases both in terms of quality and speed. First, we propose a new clustering algorithm to build “visual words” efficiently. Since our CBCD framework performs indexing using tree structures of cluster centers, the proposed clustering algorithm can improve the quality of retrieval and greatly shrink the off-line training time. Then, we introduce a TF-IDF (term frequency–inverse document frequency) weighted BOF (bag-of-features) voting retrieval method for matching video frames, which is robust to significant video distortion and efficient in terms of memory usage and computation time. Furthermore, we present a verification step robustly inspecting the temporal consistency between the query video and the corresponding candidate videos to further improve the accuracy of retrieval. The experimental results show that the proposed video copy detection scheme achieves high localization accuracy, while achieving comparable performance compared with state-of-the-art copy detection methods.