Abstract

This paper describes an original approach which jointly addresses two fundamental issues of video partitioning which represent the early important stage of any content-based video indexing system. These two issues are the detection of shot changes, and the labeling of the shot configuration related to the camera movement in terms of static shot, panning, traveling, zooming,... They are both derived from the computation, at each time instant, of the dominant motion in the image represented by a 2D affine model, and from the variation of the size of its associated support. The successive steps of the method rely on statistical techniques ensuring robustness and efficiency. In particular, it can cope with scenes containing moving objects. Results on a real documentary video are reported and validate the proposed approach.

Citations