Page 72 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience
P. 72

Section 3.4.  The H.263 Standard                               49

            3.4  The H.263 Standard


            3.4.1  Introduction
            The  H.263  recommendation  speci/es  a  coded  representation  that  can  be  used
            for compressing the moving picture component of audiovisual services at low
            bit  rates.  The  recommendation  fully  speci/es  the  decoder  and  the  bitstream
            syntax but does not explicitly specify the encoder. As already mentioned, this
            gives manufacturers a large degree of freedom to propose their own optimized
            encoders,  as  long  as  the  output  bitstream  conforms  to  the  standard  decodable
            syntax.  However,  during  the  standardization  process,  a  software-based  codec
            (encoder-decoder)  called  the  test  model  is  developed  to  study  the  core  ele-
            ments  of  the  standard.  For  example,  version  5  of  the  test  model  near-term
            (TMN) is described in Ref. 76.

            3.4.2  Source Format

            The  standard  supports  all  /ve  members  of  the  CIF  family  described  in
            Section 2.4.4 and Table 2.1. As a minimum requirement, all decoders shall be
            able to operate with SQCIF and QCIF. Encoders, on the other hand, shall be
            able to operate with either SQCIF or QCIF and are not obliged to be able to
            operate with both.

            3.4.3  Video Source Coding Algorithm

            The  generalized  form  of  the  source  coder  is  illustrated  in  Figure  3.1.
            It  is  a  hybrid  of  interpicture  prediction  to  utilize  temporal  redundancy  and
            transform coding of  the error  signal  to reduce  spatial redundancy.

            3.4.3.1  Picture Coding Structure

            The  input  video  consists  of  a  sequence  of  pictures  (or  frames).  Each  picture
            is  divided  into  groups  of  blocks  (GOBs).  A  GOB  consists  of  k × 16  lines,
            depending on the picture format (k = 1 for SQCIF, QCIF, and CIF, k = 2 for
            4CIF, and k = 4 for 16CIF). For example, there are 9 GOBs in a QCIF picture.
            Each  GOB  is  divided  into  macroblocks  (MBs).  A  macroblock  consists  of

            16 × 16 samples of Y and the spatially corresponding 8 × 8 samples of C B    and


            C R    . If we de/ne a block as 8 × 8 samples of Y ;C ,or  C , then a macroblock

                                                            R
                                                      B
            consists of 6 blocks: 4 luma blocks and the 2 spatially corresponding chroma
            blocks.  Figure  3.2  illustrates  the  H.263  picture  structure  for  a  QCIF  frame.
            As shown, GOBs are coded from top to bottom in increasing number. Within
            each  GOB,  the  MBs  are  coded  from  left  to  right  (and  from  top  to  bottom  if
            the  GOB  contains  more  than  one  row  of  MBs)  in  increasing  number.  Within
   67   68   69   70   71   72   73   74   75   76   77