Towards Viewing Quality Optimized Video Adaptation

Jingteng Xue,Chang Wen Chen

DOI: https://doi.org/10.1109/icme.2011.6012079

2011-01-01

Abstract:For video delivery over heterogeneous networks, the bandwidth mismatch at the receiver edge often leads to inevitable degradation of video quality during adaptation. While most studies in this field are based on maximizing PSNR (peak-signal-to-noise-ratio), many psychophysical experiments discover that viewing quality is not correlated well with PSNR but significantly influenced by the viewing conditions (e.g. display size and viewing distance), which are determined by the receiver device. In this research, we show that major picture quality evaluation schemes are not suitable for subjective quality driven video adaptation because they are unable to track the correspondence between viewing condition and viewing quality. In this paper we investigate a novel paradigm that performs adaptation with respect to the target display scenario in order to maximize the viewing quality. We demonstrate that viewing ratio (VR) plays a key role in linking viewing conditions and perceptual quality of given video signals. We convert the linkage between VR and perceptual quality into a computationally feasible algorithm for video adaptation. Experimental results show the VR based video adaptation scheme not only outperforms the benchmarks in terms of quality-of-experience for designated viewing scenarios, but also provides generally an improved visual quality for mobile devices with bandwidth constraint.