Fast integer-pel motion estimation on statistical properties for H.264/AVC

  • Authors:
  • Jin-Young Noh;Min-Cheol Hong

  • Affiliations:
  • Soongsil Univ. Sangdo-Ro, Dongjak-Gu, Seoul, Korea;Soongsil Univ. Sangdo-Ro, Dongjak-Gu, Seoul, Korea

  • Venue:
  • Proceedings of the 2012 ACM Research in Applied Computation Symposium
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we propose simple and efficient fast integer-pel motion estimation for H.264/AVC. In a sequence, motion vectors of neighbor blocks and the current block have correlations that are different depending on the motion of the sequence. Using the statistical properties of the motion vector and the properties of the sequence, the proposed algorithm adaptively determines a search range. Here, in the simulation result, the motion estimation time was reduced by an average of 96.2% over the full search algorithm (FS) and up to 26.4% over enhanced predictive zonal search (EPZS) at all CIF sequence. The number of search points was also reduced by an average of 29.0%, up to 57.8% over EPZS. On the other hand, BDPSNR was reduced by an average of 0.07 dB and BDBR increased by an average of 1.3% over FS.