TV commercials on broadcast television stations are repeats with essentially no changes. They can vary from several seconds in length to several minutes. The problem of finding repeating unknown commercials is highly related to video processing techniques, such as video segmentation, indexing and retrieval. In this paper, we propose a recognition system which shows how repeating commercials can be identified in a TV stream using a novel scheme. Our architecture assumes no a prior knowledge of TV streams or commercials and doesn’t require a pre-trained database. The experimental result on the live TV streams show that the proposed algorithms are effective in detecting repeats in digital television video stream.
5th International Conference on Image and Graphics (ICIG 2009). Proceedings of the 5th International Conference on Image and Graphics: ICIG 2009 (Xian, China 20-23 September, 2009) p. 863-868