In this work, we have presented a video database indexing methodology that works well for a content based video copy detection (CBVCD) system. Video data is first segmented into cohesive units called shots. A clustering based method is proposed to extract one or more Representative frames from the shots. On such collection of representatives extracted from all the shots in the video database, triangle inequality based image database indexing scheme is applied. Thus, video indexing is mapped to the task of image indexing. For a shot, following the proposed methodology primarily candidate shots corresponding to the matched representative frames are retrieved. Only on such small number of candidates the rigorous video sequence matching technique can be applied to make final decision by the CBVCD system or video retrieval system. Experimental result with a CBVCD system indicates significant gain in terms of speed, reduces false alarm rate without much compromise in terms of correct recognition rate in comparison to exhaustive search. © Springer International Publishing Switzerland 2014.
CITATION STYLE
Dutta, D., Saha, S. K., & Chanda, B. (2014). Indexing video database for a CBVCD system. In Smart Innovation, Systems and Technologies (Vol. 27, pp. 301–308). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-319-07353-8_36
Mendeley helps you to discover research relevant for your work.