We present an approach to describe moving pictures in terms of their structural properties for video coding, video haudling, and video indexing. The description contains 2D shape, motion, spatial relation, and relative depth of each region. To obtain the deseription, we develop the incremental segmentation scheme. The scheme has been desigued aloug the analysis-by-synthesis approach, and uses a sequence of images to estimate object boundaries and motion information incrementally/successively. By combining the information from extended (longer) image sequences, aud also by treating the segmentation and dynamic occlusion analysis simultaneously, the scheme attenipt to improve successively over time the accuracy of the object boundary and motion information. Some preliminary results are shown.
抄録全体を表示