Rate-distortion optimized low-delay 3D video communications

Abstract

This paper focuses on the rate-distortion optimization of low-delay 3D video communications based on the latest H.264/MVC video coding standard. The first part of the work proposes a new low-complexity model for distortion estimation suitable for low-delay stereoscopic video communication scenarios such as 3D videoconferencing. The distortion introduced by the loss of a given frame is investigated and a model is designed in order to accurately estimate the impact that the loss of each frame would have on future frames. The model is then employed in a rate-distortion optimized framework for video communications over a generic QoS-enabled network. Simulations results show consistent performance gains, up to 1.7 dB PSNR, with respect to a traditional a priori technique based on frame dependency information only. Moreover, the performance is shown to be consistently close to the one of the prescient technique that has perfect knowledge of the distortion characteristics of future frames.