• DocumentCode
    3403465
  • Title

    A dynamic slice control scheme for slice-parallel video encoding

  • Author

    Lili Zhao ; Jizheng Xu ; You Zhou ; Mingjing Ai

  • Author_Institution
    State Key Lab. of Virtual Reality Technol. & Syst., Beihang Univ., Beijing, China
  • fYear
    2012
  • fDate
    Sept. 30 2012-Oct. 3 2012
  • Firstpage
    713
  • Lastpage
    716
  • Abstract
    Slice-parallel video coding with a fixed number of slices and uniformly allocated coding regions would cause extra waiting time among slice threads, especially when the encoding machine is also busy running other applications. In this paper, we propose to adaptively decide the number of slices before encoding each frame by computing the expected value of the coding time of the encoder under two kinds of machine states: busy and free based on a Markov system. We also manage to balance computational complexity among slices by utilizing the estimated encoding time of each allocation unit in the frame to be encoded. Experimental results show that the proposed scheme can further improve the coding speed of an H.264 based fixed-slice parallel encoder by 30% on average, while the rate-distortion performance change is marginal.
  • Keywords
    Markov processes; computational complexity; multi-threading; video coding; H.264 standard; Markov system; allocation unit; busy state; coding speed improvement; coding time; computational complexity; dynamic slice control scheme; encoding machine; fixed-slice parallel encoder; free state; machine states; rate-distortion performance; slice threads; slice-parallel video coding; slice-parallel video encoding; uniform coding region allocation; Central Processing Unit; Computational complexity; Encoding; Parallel processing; Resource management; Video coding; H.264; Markov system; Slice-parallel; video coding;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Image Processing (ICIP), 2012 19th IEEE International Conference on
  • Conference_Location
    Orlando, FL
  • ISSN
    1522-4880
  • Print_ISBN
    978-1-4673-2534-9
  • Electronic_ISBN
    1522-4880
  • Type

    conf

  • DOI
    10.1109/ICIP.2012.6466959
  • Filename
    6466959