• DocumentCode
    2360286
  • Title

    Compound Block Algorithm of Moving Object and Background in H.264

  • Author

    Zou, Jianfeng ; Rong, Ding

  • Author_Institution
    State Key Lab. of Software Dev. Environ., Beihang Univ., Beijing, China
  • fYear
    2009
  • fDate
    25-27 Aug. 2009
  • Firstpage
    1604
  • Lastpage
    1606
  • Abstract
    H.264 is the newest video encoding algorithm introduced by ITU, which adopts block matching algorithm (BMA) and supports multiple reference frame. If video sequences only have single moving object and background pictures, using BMA may lead to prediction error and even decreased coding efficiency. And in the object based coding, additional bits for encoding the shape of the objects are required and syntax of H.264 is also needed to improve. In this paper, a compound block algorithm of moving object and background (CBA) is proposed. Firstly each block is regarded as a composition of a background plane and a moving object plane, and the two plane are predicted respectively. Secondly, the boundary separating the two parts is predicted from boundary in the previous decoded frame and the best vector is selected out by a boundary estimation algorithm. Finally, the prediction errors of the two parts are transmitted by transform, quantization and binary coding, and no additional bits are required. The experimental result show that the PSNR is enhanced by 0.2~0.9 db.
  • Keywords
    image motion analysis; image sequences; quantisation (signal); video coding; H.264; ITU; background pictures; binary coding; boundary estimation algorithm; compound block algorithm; moving object; object based coding; quantization; video encoding algorithm; video sequences; Decoding; Encoding; MPEG 4 Standard; Partitioning algorithms; Programming; Shape; Software algorithms; Video coding; Video sequences; Videoconference; H.264; block matching algorithm; compound block algorithm;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    INC, IMS and IDC, 2009. NCM '09. Fifth International Joint Conference on
  • Conference_Location
    Seoul
  • Print_ISBN
    978-1-4244-5209-5
  • Electronic_ISBN
    978-0-7695-3769-6
  • Type

    conf

  • DOI
    10.1109/NCM.2009.282
  • Filename
    5331436