Page 164 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience
P. 164

Chapter 6


            Multiple-Reference Motion

            Estimation Techniques





            6.1  Overview


            To  achieve  high  coding  e ciency,  Chapter  5  investigated  an  advanced  mo-
            tion estimation technique called warping-based motion estimation. This chapter
            considers  another  advanced  technique,  called  multiple-reference  motion  esti-
            mation.
               In  multiple-reference  motion-compensated  prediction  (MR-MCP),  motion
            estimation  and  compensation  are  extended  to  utilize  more  than  one  reference
            frame. The reference frames are assembled in a multiframe memory (or bu&er)
            that  is  maintained  simultaneously  at  encoder  and  decoder.  In  this  case,  in
            addition  to  the  spatial  displacements,  a  motion  vector  is  extended  to  also
            include a temporal  displacement.
               This  chapter  investigates  the  prediction  gains  achieved  by  MR-MCP.  Par-
            ticular  emphasis  is  given  to  coding  e ciency  at  very  low  bit  rates.  More
            precisely, the chapter attempts to answer the following question: Is the use of
            additional bit rate to transmit the extra temporal displacement justi able in
            terms of an improved rate-distortion performance? The chapter also examines
            the properties of the multiple-reference block-motion )eld and compares them
            to those of  the single-reference  case.
               The  rest  of  the  chapter  is  organized  as  follows.  Section  6.2  brie-y  re-
            views  multiple-reference  motion  estimation  techniques.  Section  6.3  concen-
            trates  on  the  long-term  memory  multiple-reference  motion  estimation  tech-
            nique.  The  section  starts  by  examining  the  properties  of  multiple-reference
            block-motion  )elds  and  compares  them  to  those  of  single-reference  )elds.  It
            then investigates the prediction gains and the e ciency of the long-term mem-
            ory technique at very low bit rates. The chapter concludes with a discussion in
            Section 6.4.

                                           141
   159   160   161   162   163   164   165   166   167   168   169