• DocumentCode
    318274
  • Title

    Block matching motion estimation using block integration based on reliability metric

  • Author

    Yoshida, Toshiyuki ; Katoh, Hidetoshi ; Sakai, Yoshinori

  • Author_Institution
    Imaging Sci. & Eng. Lab., Tokyo Inst. of Technol., Japan
  • Volume
    2
  • fYear
    1997
  • fDate
    26-29 Oct 1997
  • Firstpage
    152
  • Abstract
    This paper proposes a motion estimation technique to improve accuracy and coherence of motion vectors estimated by the standard block matching technique. The proposed technique is based on a matching block integration using a reliability metric of motion vectors. First, the reliability metric is computed for all the matching blocks in the target image and then, based on this value and its dependency on the direction, four neighboring matching blocks are checked whether they should be integrated or not. Finally, motion vectors are estimated for the integrated matching blocks by the block matching technique. Since the block integration is performed only for flat matching blocks or those with a simple edge, the coherence is improved while the performance of motion compensation is kept as far as possible. The examples given show the effectiveness of the proposed approach
  • Keywords
    edge detection; image classification; image matching; image sequences; motion estimation; reliability theory; video coding; block integration; block matching technique; coherence; flat matching blocks; motion compensation; motion estimation; motion vectors; neighboring matching blocks; reliability metric; simple edge blocks; target image; Bit rate; Estimation error; Fluctuations; Image edge detection; Motion analysis; Motion compensation; Motion estimation; Redundancy; Video coding;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Image Processing, 1997. Proceedings., International Conference on
  • Conference_Location
    Santa Barbara, CA
  • Print_ISBN
    0-8186-8183-7
  • Type

    conf

  • DOI
    10.1109/ICIP.1997.638696
  • Filename
    638696