Enhanced Mode-Adaptive Fine Granularity Scalability

Wen-Hsiao Peng*, Yen Kuang Chen

*Corresponding author for this work

Research output: Contribution to journalArticle

2 Scopus citations

Abstract

In this article, we present a scalable video compression algorithm to deliver higher compression efficiency with limited drifting error. MPEG-4 Fine Granularity Scalability (FGS) compresses the video into a base layer and an enhancement layer. Currently, because the enhancement layer is predicted from the poor-quality base layer, the compression efficiency is low. To improve the compression efficiency, we construct enhancement-layer predictors from (1) macroblocks of current reconstructed base-layer frame, (2) macroblocks of previously reconstructed enhancement-layer frame, and (3) the average of previous two. On the other hand, the unpredictable receiving manner of enhancement layer could cause predictor mismatch error. The predictor mismatch error further results in drifting error. To minimize the drifting error, we create an adaptive mode-selection algorithm, in the encoder, which first smartly estimates possible drifting error of the decoder side and then uses the best macroblock modes wisely. In this article, we show that predictors constructed jointly from the base-layer frame and the enhancement-layer frame can reduce the drifting error. And, predictors constructed from the base-layer frame can stop the drifting error. As compared to other advance FGS schemes, our algorithm shows 0.3-0.5 dB PSNR improvement with a less complex structure. Although compared to MPEG-4 FGS, more than 1-1.5 dB quality improvement can be gained.

Original languageEnglish
Pages (from-to)308-321
Number of pages14
JournalInternational Journal of Imaging Systems and Technology
Volume13
Issue number6
DOIs
StatePublished - 1 Dec 2003

Keywords

  • Fine granularity scalability
  • Layered video coding
  • Streaming video

Fingerprint Dive into the research topics of 'Enhanced Mode-Adaptive Fine Granularity Scalability'. Together they form a unique fingerprint.

  • Cite this