Dynamic threshold method for scene change detection
Sungju-Youm, Woosaeng-Kim · 2003
In general, the scene change for video indexing could be detected by using an appropriate threshold. However, the single fixed threshold value used in previous researches cannot perform well for all videos equally. In this paper, we propose a dynamic threshold method (DTM) in order to improve the performance of the scene change detection. The proposed method utilizes variable parameters obtained from detected cuts and the distribution of cut intervals in order to change the threshold dynamically. In short, the threshold in DTM is lowered in the area where the chance of cut occurrence is high and raised in the area where the chance of cut occurrence is low. Experimental results indicate that the proposed DTM is both computationally low and robust to false alarm.