An Objective Metric for Stereoscopic 3D Video Quality Prediction Using Perceptual Thresholds

摘要

This study proposes an objective metric dedicated to stereoscopic 3D video quality assessment, which characterizes a detected problem based on its influence on viewers' perception and avoids a direct prediction of the mean opinion score. Once the problem (degradation or artifact) is detected, it can be automatically classified into one of the following color categories: (1) Green, for no problems perceived; (2) Orange, which indicates the problem is acceptable but induces visual annoyance; or (3) Red, for unacceptable problem level. The boundary between the Green and the Orange categories defines the visual annoyance threshold, while the boundary between the Orange and the Red categories defines the acceptability threshold. This metric can be adapted to service or production requirements by changing the percentage of acceptability.