Organisation internationale de normalisation

5.6Mixed-resolution MVC

5.6.1.1.1JVT-Z026 ( Prop 2.2.1/3.1) [Y. Chen (TUT), Y.-K. Wang (USTC), S. Liu, M. M. Hannuksela (Nokia), H. Li (Nokia)] On asymmetric MVC

This is a follow-up proposal to JVT-Y054. "Asymmetric coding" involves the coding of two views of a stereoscopic video with different resolutions. In JVT-Y054, it was asserted that the proposed scheme has low complexity and requires a smaller decoded picture buffer size compared to downsampled inter-view prediction. Simulation results under common test condition were provided. In this proposal, results on stereoscopic video (two views) were provided. It was claimed that the proposed method has almost the same efficiency as downsampled inter-view prediction.
Rather than downsampling, uses 1/2 sample interpolation as 1/4 sample positions and uses odd integer samples as 1/2 sample positions, and even integer samples as integer-sample positions.
Results were compared to simulcast and to the downsampled reference technique. About the same performance, relative to downsampled reference technique, was reported (a small gain), and about 14% gain relative to simulcast-resolution use of MVC design.
Software? Can be made available.
Draft text? Not yet.
JVT Disposition: Further study was encouraged.

5.6.1.1.2JVT-Z034 ( Prop 2.2) [S. Cho, B. Lee, N. Hur, J. Kim, S.-I. Lee (ETRI)] Prelim subjective test results for mixed resolution stereo video coding

This document reported preliminary results of subjective tests for mixed resolution stereo video coding. Such coding methods were envisaged to be used for 3D DMB services and systems in the future. The presented results were partly achieved in collaboration with Fraunhofer HHI. They were reported to be considered to be preliminary and to report work in progress. Test conditions were reported to be "not yet perfect" and the results were reported to be not consistent and complete. So far the results reportedly do not allow drawing reliable conclusions. However, there was reported to be some indication that mixed resolution stereo coding may outperform MVC at least in some cases, under some conditions.
Shown on 3.5 inch display, each view having 320x480 resolution, 30 fps.
3 test sequences.
CAVLC, no 8x8 transform.
Results were characterized as preliminary.
Conclusions difficult – not consistent The author indicated an interest in providing improved measurements in the future.
Disposition: Further study was encouraged.