Page 108 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience
P. 108

Section 3.5.  The MPEG-4 Standard                              85


            segment.  For  example,  a  background  sprite  (also  referred  to  in  the  literature
            as  a  background  mosaic)  can  be  constructed  by  collecting  all  pels  belong-
            ing  to  the  background  throughout  a  video  segment.  Note  that  in  the  case  of
            camera  panning,  for  example,  the  background  sprite  can  be  larger  than  the
            actual  frames  of  the  sequence.  This  still,  and  possibly  large,  image  needs  to
            be transmitted only once before transmitting the corresponding video segment.
            For  each  frame  of  the  video  segment,  there  is  no  need  to  encode  a  back-
            ground  VOP.  Instead,  a  small  number  of  parameters  needs  to  be  transmitted
            to allow the decoder to warp=crop the sprite and generate an appropriate back-
            ground VOP. Thus, in such cases, sprite coding can achieve very high coding
            e,ciency.
               Sprite  coding  can  operate  in  three  modes:  basic  sprite  coding,  low-latency
            sprite coding, and scalable sprite coding. In basic sprite coding the whole sprite
            is encoded and transmitted to the decoder before transmitting the corresponding
            video segment. In low-latency sprite coding only part of the sprite is encoded
            and  transmitted.  This  part  is  su,cient  to  be  used  for  the  /rst  few  frames  of
            the video segment. The remaining part of the sprite is transmitted, piecewise,
            when required or as the bandwidth allows. In scalable sprite coding the sprite
            is encoded and transmitted progressively. In other words, a low-quality version
            of the sprite is encoded and transmitted /rst. This is then re/ned gradually by
            encoding and transmitting residuals.

            3.5.7  Scalability

            MPEG-4  supports  both  temporal  and  spatial  scalability  using  multiple  VOLs.
            For  example,  in  the  case  of  two  VOLs,  one  VOL  provides  the  base  layer
            whereas the other  VOL  provides the enhancement layer.
               MPEG4 uses a generalized scalability framework, as shown in Figure 3.15.
            In this framework the functionality of a block depends on the chosen type of
            scalability.


                         I                                      O
                         1    Enhancement-            Enhancement-  1
                              Layer Encoder          Layer Decoder
                                                 Demultiplexer  Midprocessor
                In                                                      Out 1
                    Scalability                                  Scalability
                              Midprocessor
                   Preprocessor                                 Postprocessor   Out 0
                                         Multiplexer
                               Base-Layer             Base-Layer
                               Encoder                 Decoder
                         I                                      O
                         0                                       0
                              Figure 3.15:  MPEG-4 generalized  scalability
   103   104   105   106   107   108   109   110   111   112   113