• DocumentCode
    2708443
  • Title

    Recent advances in object-based image compression

  • Author

    Schmalz, Mark S.

  • Author_Institution
    Center for Comput. Vision Res., Florida Univ., Gainesville, FL, USA
  • fYear
    2005
  • fDate
    29-31 March 2005
  • Firstpage
    478
  • Abstract
    Summary form only given. Image compression has progressed from pixel-level entropy coding to transform coding applied to fixed regions (e.g., JPEG-2K with rectangular blocks). Our interest is in substitutions applied to arbitrarily-shaped regions. For example, object-based compression (OBC) segments disjoint, contiguous image regions, and compactly approximates each region´s boundary and content (e.g., represented by a codebook exemplar). When region boundaries closely approximate scene object boundaries, OBC resembles object recognition. Disadvantages include cost of segmentation and region boundary representation, and pixel- or object-level distortion in the decompressed image.
  • Keywords
    data compression; distortion; image coding; image representation; image segmentation; object recognition; JPEG-2K; codebook exemplar; decompressed image object-level distortion; decompressed image pixel-level distortion; image segmentation; object recognition; object-based image compression; pixel-level entropy coding; region boundary representation; scene object boundaries; transform coding; Chromium; Computer vision; Entropy coding; Image coding; Image segmentation; Layout; Object recognition; Pixel; Transform coding; Visualization;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Data Compression Conference, 2005. Proceedings. DCC 2005
  • ISSN
    1068-0314
  • Print_ISBN
    0-7695-2309-9
  • Type

    conf

  • DOI
    10.1109/DCC.2005.79
  • Filename
    1402235