Towards efficient copy detection for digital videos by using spatial and temporal features

This research aims at developing a practical content-based video copy detection scheme, which employs an efficient and effective content matching approach to help determine whether an investigated video is a duplicated copy and may infringe the copyright. In the proposed scheme, the key frames of videos are first retrieved and the methods of Vector Quantization (VQ) and Singular Value Decomposition (SVD) will be applied to extract the spatial features of these frames. After locating the visually similar key-frame pairs by the matched VQ codewords, we use the shot lengths as the temporal features for further matching to achieve a more accurate result. The experimental results will show the feasibility of this potential solution.