• DocumentCode
    474383
  • Title

    Multiple Background Sprite Generation Using Camera Motion Characterization for Object-Based Video Coding

  • Author

    Krutz, Andreas ; Glantz, Alexander ; Haller, Martin ; Droese, Michael ; Sikora, Thomas

  • Author_Institution
    Commun. Syst. Group, Tech. Univ. Berlin, Berlin
  • fYear
    2008
  • fDate
    28-30 May 2008
  • Firstpage
    313
  • Lastpage
    316
  • Abstract
    Recent work has shown that object-based video coding can provide higher coding gain than common H.264/AVC for single-view and the MVC standard based on H.264 for multi-view (MVC). The use of background sprites outperformes the AVC/MVC especially in sequences containing rotating camera motion and moving foreground objects. The coding performance strongly relies on the preprocessing steps, e.g. sprite generation and object segmentation. In this paper, we present an enhanced background sprite generation algorithm for object-based single- and multi-view video coding (OBVC/OBMVC). It is a new feature for our OBVC/OBMVC recently proposed. We produce multiple background sprites based on camera motion characterization and physical camera parameter estimation. Experimental results show how these multiple sprites increase the coding performance for single- and multi-view sequences.
  • Keywords
    estimation theory; feature extraction; image motion analysis; image sequences; video cameras; video coding; camera motion characterization; feature extrcation; image sequence; multiple background sprite generation; object-based video coding; physical camera parameter estimation; Automatic voltage control; Cameras; Character generation; Codecs; Motion analysis; Motion estimation; Object segmentation; Parameter estimation; Sprites (computer); Video coding; H.264/AVC; camera motion analysis; multiview video coding; sprite generation;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    3DTV Conference: The True Vision - Capture, Transmission and Display of 3D Video, 2008
  • Conference_Location
    Istanbul
  • Print_ISBN
    978-1-4244-1760-5
  • Electronic_ISBN
    978-1-4244-1755-1
  • Type

    conf

  • DOI
    10.1109/3DTV.2008.4547871
  • Filename
    4547871