• DocumentCode
    885447
  • Title

    Subband HDTV coding using high-order conditional statistics

  • Author

    Lei, Shaw-Min ; Chen, Ting-Chung ; Tzou, Kou-Hu

  • Author_Institution
    Bellcore, Red Bank, NJ, USA
  • Volume
    11
  • Issue
    1
  • fYear
    1993
  • fDate
    1/1/1993 12:00:00 AM
  • Firstpage
    65
  • Lastpage
    76
  • Abstract
    An HDTV coding scheme that is capable of achieving high quality below 20 Mb/s is presented. The system employs motion estimation/compensation followed by four-band subband splitting. The band signals are then subject to quantization and coded by an innovative high-order entropy coder. It is shown that this high-order entropy coder generates a bit rate nearly 30% lower than that of run-length/Huffman coding for motion-compensated frame differences. About 15% improvement is observed for the intrafield coded data. To reduce the complexity of the high-order entropy coder for high-speed implementation, two techniques, codebook reduction and nonuniform quantization of conditioning samples, are developed. A method for selecting conditioning samples and their order, based on mutual information between samples, is proposed and shown to outperform several other systematic and intuitive approaches. Simulations show that the new system, at about 20 Mb/s, achieves picture quality comparable to that of a 120-Mb/s six-band intrafield coding system previously investigated
  • Keywords
    high definition television; image coding; motion estimation; statistical analysis; vector quantisation; 20 Mbit/s; VQ; codebook reduction; conditioning samples; four-band subband splitting; high-order conditional statistics; high-order entropy coder; intrafield coded data; motion estimation/compensation; nonuniform quantization; subband HDTV coding; B-ISDN; Bit rate; Entropy coding; HDTV; Local area networks; Motion estimation; Mutual information; Optical fiber LAN; Quantization; Statistics;
  • fLanguage
    English
  • Journal_Title
    Selected Areas in Communications, IEEE Journal on
  • Publisher
    ieee
  • ISSN
    0733-8716
  • Type

    jour

  • DOI
    10.1109/49.210545
  • Filename
    210545