Adaptive motion search algorithms for fast H.264 inter prediction

Shih Tse Wei, Chun Yung Wen, Bin Da Liu, Jar Ferr Yang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

The adaptive motion search algorithm and its corresponding architecture for H.264/AVC are proposed in this paper. The correlations of spatial MBs and temporal MB are exploited to predict the search centers. The origin point is chosen as another search center to enhance the motion search in case of prediction error. Moreover, according to the spatial and temporal correlations, three kinds of search patterns are used to reduce the search points. The average number of search points in the proposed algorithm is about 13 and the performance is similar to the full search algorithm. To support the irregular search patterns, a hardware architecture design is also proposed. Synthesis result shows that the total number of gate counts is 21.7k and the maximum operation frequency is 142 MHz. It can achieve the requirement for real-time encoding of 720p-HD video format.

Original languageEnglish
Title of host publication2008 International Conference on Communications, Circuits and Systems Proceedings, ICCCAS 2008
Pages689-692
Number of pages4
DOIs
Publication statusPublished - 2008 Dec 1
Event2008 International Conference on Communications, Circuits and Systems, ICCCAS 2008 - Xiamen, Fujian Province, China
Duration: 2008 May 252008 May 27

Publication series

Name2008 International Conference on Communications, Circuits and Systems Proceedings, ICCCAS 2008

Other

Other2008 International Conference on Communications, Circuits and Systems, ICCCAS 2008
Country/TerritoryChina
CityXiamen, Fujian Province
Period08-05-2508-05-27

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications
  • Hardware and Architecture
  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Adaptive motion search algorithms for fast H.264 inter prediction'. Together they form a unique fingerprint.

Cite this