Prediction of the Influence of Navigation Scan-path on Perceived Quality of Free-Viewpoint Videos

摘要: Free-Viewpoint Video (FVV) systems allow the viewers to freely change viewpoints of scene. In such systems, view synthesis and compression are two main sources artifacts influencing perceived quality. To assess this influence, quality evaluation studies often carried out using conventional displays generating predefined navigation trajectories mimicking possible movement when exploring content. Nevertheless, as different may lead conclusions in terms visual benchmarking performance methods identify critical needed. This paper aims at impact exploration (defined Hypothetical Rendering Trajectories: HRT) on FVV subjectively objectively, providing contributions. Firstly, a subjective assessment test including HRTs was analyzed. The results demonstrate quantify influence HRT Secondly, we propose new objective video measure objectively predict HRT. measure, based Sketch-Token representation, models how categories contours spatially temporally from higher semantic level. Performance comparison with existing metrics for FVV, highlight promising automatic detection most benchmark immersive systems.