• DocumentCode
    2289146
  • Title

    Extended Macroblock Bipartitioning Modes for H.264/AVC Inter Coding

  • Author

    Vermeirsch, Kenneth ; Cock, Jan De ; Notebaert, Stijn ; Lambert, Peter ; Van de Walle, Rik

  • Author_Institution
    Dept. of Electron. & Inf. Syst., Ghent Univ., Ghent
  • fYear
    2008
  • fDate
    15-17 Dec. 2008
  • Firstpage
    142
  • Lastpage
    147
  • Abstract
    We present a family of new macroblock partitions for H.264/AVC inter prediction. These modes allow a macroblock to be bipartitioned along a horizontal, vertical, or diagonal edge with an arbitrary location within the block. These new partitioning modes allow an encoder to adapt more closely to the local motion characteristics and improve its rate-distortion performance. In order to quantify the improvements we implemented the proposed partitioning modes in the context of the H.264/AVC video coding standard. Over half the coding efficiency of existing more general schemes is attained at a fraction of the number of added partitioning modes, and hence of encoder complexity, of those schemes. The results are consistent across different profiles, indicating that the proposed extension has good compatibility with the existing H.264/AVC tools for motion compensation and residual coding.
  • Keywords
    block codes; computational complexity; motion compensation; rate distortion theory; video coding; H.264/AVC inter coding; H.264/AVC video coding standard; encoder complexity; extended macroblock bipartitioning mode; local motion characteristics; motion compensation; rate-distortion performance; residual coding; Automatic voltage control; Decoding; IEC standards; ISO standards; Information systems; Multimedia systems; Shape; Streaming media; Video coding; Video compression; AVC; H.264; coding efficiency; motion compensation; partitioning;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Multimedia, 2008. ISM 2008. Tenth IEEE International Symposium on
  • Conference_Location
    Berkeley, CA
  • Print_ISBN
    978-0-7695-3454-1
  • Electronic_ISBN
    978-0-7695-3454-1
  • Type

    conf

  • DOI
    10.1109/ISM.2008.58
  • Filename
    4741160