View upsampling optimization for mixed resolution 3D video coding

Michal Joachimiak, Miska M. Hannuksela, Moncef Gabbouj

    Research output: Contribution to journalArticleScientificpeer-review

    1 Citation (Scopus)
    11 Downloads (Pure)


    3D video is composed out of two or more, temporally synchronized, 2D video streams acquired at different camera poses and accompanied by geometrical information. In a mixed resolution 3D video stream, a subset of views is coded at reduced resolution. It has been shown in the literature that subjective quality of mixed resolution 3D video is close to that of full resolution 3D video. In order to improve the coding gain in mixed resolution coding scenario we present a new depth encoding method called view upsampling optimization. A novel depth distortion metric based on the performance of the depth-based super resolution is also presented. Finally, to improve the quality of the decoded video an improved depth-based super resolution method that uses view synthesis quality mapping is used for upsampling of low resolution views. The simulations, performed with the recently standardized MVC+D encoder, show that the proposed solution combined with the state of the art view synthesis distortion outperforms the anchor MVC+D coding scheme by 14.5 % of dBR on average for the total coded bitrate and by 17 % of dBR on average for the synthesized views.
    Original languageEnglish
    JournalMultidimensional Systems and Signal Processing
    Publication statusPublished - 2015
    Publication typeA1 Journal article-refereed

    Publication forum classification

    • Publication forum level 1


    Dive into the research topics of 'View upsampling optimization for mixed resolution 3D video coding'. Together they form a unique fingerprint.

    Cite this