Two-dimensional motion vector coding for low bit rate videophone applications

Most of the current video coding standards that use motion compensation for interframe redundancy reduction provide a 1-dimensional VLC (variable length coding) scheme to separately encode the horizontal and vertical components of the differential motion vectors (DMVs). It can be derived, however, from information theory that coding the DMV components separately is less efficient than coding them jointly as a single entity. In this report, we propose a two-dimensional VLC scheme to code the differential motion vectors for videophone applications. Experimental results based on five commonly used video test sequences show that a 24% average saving of motion vector bits, as compared to the 1-D scheme in H.263, has been achieved without sacrificing picture quality at all