Parametric overlapped block motion compensation for pixel-adaptive temporal prediction

Yi Wen Chen*, Chung Lin Lee, Wen-Hsiao Peng

*Corresponding author for this work

    Research output: Contribution to conferencePaperpeer-review

    1 Scopus citations

    Abstract

    In this paper, a Parametric Overlapped Block Motion Compensation (POBMC) technique is proposed to improve the temporal prediction. It extends the notion of OBMC in H.263 to accommodate the variable block-size motion partition in H.264/AVC. The approach solves for the optimal weights of OBMC as closed-form formulas involving only the distances between the predicted pixel and its nearby block centers. On average, the single use of POBMC has 4.62% BDBR saving and 0.15dB BDPSNR gain over the H.264/AVC. Furthermore, when combined with Enhanced Adaptive Interpolation Filter (EAIF), Rate-Distortion Optimized Quantization (RDOQ), Quard-tree Based Adaptive Loop Filter (QALF), Extended Macroblock (EMB), Mode-dependent Directional Transform (MDDT), and Template Matching Prediction Skip (TMP-Skip), the POBMC is observed to have BDBR savings of 22%-41% and BDPSNR gains of 0.9dB-2dB relative to H.264/AVC anchors. Our scheme is among top 10 with low delay settings in the subjective quality evaluation of high-efficiency video coding (HEVC) in the 1st JCTVC meeting in April 2010.

    Original languageEnglish
    Pages636-642
    Number of pages7
    StatePublished - 1 Dec 2010
    Event2nd Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2010 - Biopolis, Singapore
    Duration: 14 Dec 201017 Dec 2010

    Conference

    Conference2nd Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2010
    Country/TerritorySingapore
    CityBiopolis
    Period14/12/1017/12/10

    Fingerprint

    Dive into the research topics of 'Parametric overlapped block motion compensation for pixel-adaptive temporal prediction'. Together they form a unique fingerprint.

    Cite this