• DocumentCode
    2018297
  • Title

    Comparison of winner-take-all motion compensation schemes for embedded wavelet coding

  • Author

    Ang, Li-Minn ; Cheung, Hon Nin ; Eshraghian, Kamran

  • Author_Institution
    Sch. of Eng. & Math., Edith Cowan Univ., Joondalup, WA, Australia
  • Volume
    1
  • fYear
    1999
  • fDate
    1999
  • Firstpage
    390
  • Abstract
    Presents a comparison of motion compensation (MC) schemes using winner-take-all (WTA) networks for the embedded zero-tree wavelet (EZW) algorithm. The proposed MC schemes maintain the advantages of the EZW algorithm in terms of coding efficiency and scalable coding. The approaches are based on using the ancestor-descendant tree hierarchies from the EZW algorithm to form the candidate matrices for the WTA network. The various WTA MC schemes differ from one another in the way the candidate matrices are used to update a codebook. The various WTA MC schemes are compared, and simulation results using standard video sequences are presented
  • Keywords
    image sequences; matrix algebra; motion compensation; neural nets; transform coding; tree codes; video coding; wavelet transforms; ancestor-descendant tree hierarchies; candidate matrices; codebook updating; coding efficiency; embedded wavelet coding; embedded zero-tree wavelet algorithm; motion compensation schemes; scalable coding; simulation; video sequences; winner-take-all networks; Arithmetic; Australia; Discrete wavelet transforms; Image coding; Maintenance engineering; Mathematics; Motion compensation; Scalability; Streaming media; Video sequences;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on
  • Conference_Location
    Perth, WA
  • Print_ISBN
    0-7803-5871-6
  • Type

    conf

  • DOI
    10.1109/ICONIP.1999.844020
  • Filename
    844020