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

TL;DR

This paper investigates how different navigation trajectories affect perceived quality in free-viewpoint videos, using subjective tests and a new objective measure based on Sketch-Token representation to predict critical trajectories.

Contribution

It introduces a subjective assessment of navigation trajectories' impact on perceived quality and proposes a novel objective quality metric for predicting critical trajectories in FVV.

Findings

HRT significantly influences perceived quality.

The new metric effectively predicts critical trajectories.

Promising results for automatic detection of impactful navigation paths.

Abstract

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