Multiparameter method for analysis and selection of motion estimation algorithms for video compression

Evgeny Kaminsky, Ofer Hadar

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

This paper proposes a new multiparameter method for analysis and selection of motion estimation algorithms for video compression. We present motion estimation algorithms, results of computer simulations and illustrate the analysis with tables, PSNR and performance plots. Numerous algorithms and tests for analysis of algorithm performance for video compression have recently been suggested, which has resulted in a need for effective evaluation methods. A highly qualified expert is also needed to evaluate the test results. The more input parameters used the more complex and subjective the evaluation will be. Our multiparameter method for algorithm analysis and selection eliminates subjectivity and provides a qualitative and quantitative evaluation of the tested algorithms for any number of algorithms and parameters. We propose two new methods of evaluation: (1) a quality method-a graphic method using the Pareto approach, and (2) a quantity method which obtains an integrated parameter composed of numerous evaluation parameters. In addition, we evaluate various motion estimation algorithms accordingly to two different implementation strategies: (a) using a software video encoder that depends on available processing resources using a computational complexity-rate-distortion (C-R-D) evaluation framework and (b) using a power-limited video encoder implemented on mobile or handheld computing platform by using energy-rate-distortion (E-R-D) behavior.

Original languageEnglish
Pages (from-to)119-146
Number of pages28
JournalMultimedia Tools and Applications
Volume38
Issue number1
DOIs
StatePublished - 1 May 2008

Keywords

  • Algorithm evaluation
  • Algorithm selection
  • Block Matching Algorithms (BMA)
  • Fast Motion Estimation (ME)
  • Multiparameter analysis
  • Pareto approach
  • Video analysis

Fingerprint

Dive into the research topics of 'Multiparameter method for analysis and selection of motion estimation algorithms for video compression'. Together they form a unique fingerprint.

Cite this