A sense of speed in camera motion

In this paper, we propose a new approach to determine speed of camera motion in video sequences. The technique is based on the analysis of motion trajectories of image feature. The corners and interesting points of image are first detected and then tracked along an image sequence. The spatio-temporal information of image feature is a key significance in determining camera motion and speed value. We conducted the experiments to investigate the sense of speed in camera motion. The experimental results classify magnitude of speed into three levels of motion (i.e., slow, medium, and fast). The sense of speed in camera motion helps infer higher-level semantic content and query information in video retrieval and hence facilitates efficient searching of video sequence into its temporal domain.