Parametric Quality Models for Multiscreen Video Systems
Nabajeet Barman, Rahul Vanam, Yuriy A. Reznik · 2022
We propose simple parametric models for predicting visual quality scores on different devices in multiscreen systems. As input parameters, the proposed models take the distortion measure for the encoded video and parameters of viewing setup: the resolution of projected video, size of the display, and viewing distance. We derive models for the following distortion measures: PSNR, SSIM, VIF, and VMAF. We validate the proposed models using datasets corresponding to three different reproduction environments: standard TV sets, UltraHD TV sets, and mobiles. The obtained results confirm the improved accuracy of the prediction of MOS scores by the proposed techniques. The paper also includes introductory material explaining the usefulness of parametric quality for the analysis and optimizations of multiscreen video systems.