Object-based encoding: next-generation video compression

The object-based approach to video compression has long been recognized within the MPEG community as having the potential to deliver quality video at substantially lower bit rates than that afforded by its more well established block-based cousins. This potential has been demonstrated with synthetically generated video sequences rendered from 3D graphics models, and with natural video layers segmented through chroma-key techniques or hand segmentation. This paper reports the development of a fully automatic object-based encoding system. For the first time, the benefits of object-based compression can be realized for natural video sequences without hand segmentation or special production techniques.