One-step fractional motion estimation in H.264/AVC based on early predicted candidate

Nam Thang Ta, Jun Rim Choi

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

The conventional two-step fractional motion estimation algorithm has been broadly adopted in the literature due to its high encoding performance. However it induces a huge computational complexity as well as a long latency. In this paper, we propose a fast fractional motion estimation algorithm. Based on high correlation between the motion vector of a block and its uplayer, as well as relationship of integer candidates, a onestep algorithm is proposed which not only reduces the computational complexity by eliminating unnecessary fractionalpels, but also saves hardware cost. Experimental results show that the proposed design can save 32% gate count, and reduce the latency by 39% compared with previous designs, while nearly maintaining the coding performance.

Original languageEnglish
Pages (from-to)982-987
Number of pages6
JournalIEICE Electronics Express
Volume7
Issue number14
DOIs
StatePublished - 25 Jul 2010

Keywords

  • Fractional motion estimation
  • h.264
  • Video signal processing

Fingerprint

Dive into the research topics of 'One-step fractional motion estimation in H.264/AVC based on early predicted candidate'. Together they form a unique fingerprint.

Cite this