News Video Retrieval by Learning Multimodal Semantic Information

摘要

With the explosion of multimedia data especially that of video data, requirement of efficient video retrieval has becoming more and more important. Years of TREC Video Retrieval Evaluation (TRECVID) research gives benchmark for video search task. The video data in TRECVID are mainly news video. In this paper a compound model consisting of several atom search modules, i.e., textual and visual, for news video retrieval is introduced. First, the analysis on query topics helps to improve the performance of video retrieval. Furthermore, the multimodal fusion of all atom search modules ensures to get good performance. Experimental results on TRECVID 2005 and TRECVID 2006 search tasks demonstrate the effectiveness of the proposed method.