Chapter 6 - Multiple-Reference Motion Estimation Techniques

Abstract

In multiple-reference motion-compensated prediction (MR-MCP), motion estimation and compensation are extended to utilize more than one reference frame. The main aim of MR-MCP is to improve coding efficiency. A number of MR-MCP techniques have been proposed for inclusion within MPEG-4: global motion compensation (GMC), dynamic sprites (DS), and short-term frame memory/long-term frame memory (STFM/LTFM) prediction. The main difference among the various MR-MCP techniques is the way they generate the reference frames. The simplest and least computationally complex approach is the long-term memory motion-compensated prediction (LTM-MCP) technique, where past-decoded frames are assembled in the multiframe memory. The properties of long-term block-motion fields are that t he distribution of the long-term memory spatial displacements is center-biased, the distribution of the long-term memory temporal displacement is zero-biased, and the long-term memory block-motion field is smooth and varies slowly. LTM-MCP extends the motion vector of a block by a third component, which is the temporal displacement or the index into the multiframe memory. The transmission of this extra component incurs an additional bit rate compared to the single-reference case. This additional bit rate has to be justified in terms of an improvement in the rate-distortion (R-D) performance. The chapter investigates the R-D performance of the LTM-MCP technique, with particular emphasis on the efficiency of this technique at the very low bit rates typical of mobile video communication. Four H.263-like encoders were implemented in this study: single-reference (SR) encoder, SR rate-constrained (SR-RC) encoder, multiple-reference (MR) encoder, and MR rate-constrained (MR-RC) encoder.