A Bitmap Indexing Approach for Video Signature and Copy Detection

In this paper, a framework for video signature indexing and copy detection incorporating a bitmap structure is proposed. The work we present improves the computational time during the detection process of video matching, as well as reducing the space needed to store the video signatures. In addition, the method improves the detection in the case where a caption or logo has been inserted into the original video. The proposed framework is composed of two levels of bitmap indexing. The first level reduces the time taken for performing video matching between a query video and videos in the database. This can be achieved by grouping videos (keyframes) into clusters and using them as the first level index. During the video matching process, this index is used to determine the relevant clusters, so that the video in question need only be matched with those clusters, rather than the entire database, significantly reducing the computational time. The second level index reduces the space required to store the video signatures. This is achieved by converting the video signatures into bitmap vectors. In addition, another advantage of adopting a bitmap indexing method is that a low-cost Boolean operation such as AND, OR, and XOR can be utilized in the analysis. This two-level indexing scheme is simple but efficient and improves the overall system performance.