Morphological moving object segmentation and tracking for content-based video coding

Besides contour and texture as two visual primitives [1, 2], motion is another important visual primitive. The understanding of motion in a dynamic scene represents a principal research topic for many applications such as video coding, video indexing and annotation, matching and recognition. In fact, motion information contains very rich cue to describe how a dynamic scene is constructed. Such description about the independent movement of each region or object is of primary interest to all those applications. Obviously, this compact description can be utilized to improve the coding efficiency as well as to provide the foundation for all the new functionalities which are under intensive investigation in the framework of next international standard MPEG4 [3].