Reliability of 2D quality assessment methods for synthesized views evaluation in stereoscopic viewing conditions

Abstract

This paper investigates the reliability of objective quality metrics commonly used for the quality assessment of 2D media, in the context of 3D Video. In the absence of any dedicated tool for the evaluation of synthesized views quality, we often rely on available 2D metrics for direct evaluation of 3D media quality, or with some adaptation to the 3D case. However, recent studies showed that the use of DIBR, depending on its in-painting strategy, can lead to downsides whose range in terms of quality, has never been experienced with 2D media. This paper questions the reliability of the objective quality metrics normally used for the quality assessment, when assessing stereopairs. Seven DIBR algorithms are used to generate novel viewpoints. A series of commonly used quality metrics then assess their quality. The results of our experiments showed that the metrics are not sufficient to faithfully predict human judgment. Moreover, we compared our results with an experiment run in monoscopic viewing condition and the differences are for the less unexpected since the preferred artifacts in monoscopic condition are the most rejected in stereoscopic condition. This paper proposes some explanations.