Evaluation of Stereoscopic Images: Beyond 2D Quality

Abstract

Perceived image quality is a standard evaluation concept for 2D imaging systems. When applied to stereoscopic 3D imaging systems, however, it does not incorporate the added value of stereoscopic depth. Higher level evaluation concepts (naturalness and viewing experience) are proposed that are sensitive to both image quality and stereoscopic depth. A 3D Quality Model is constructed in which such higher level evaluation concepts are expressed as a weighted sum of image quality and perceived depth. This model is validated by means of three experiments, in which stereoscopic depth (camera base distances and screen disparity) and image quality (white Gaussian noise and Gaussian blur) are varied. The resulting stimuli are evaluated in terms of naturalness, viewing experience, image quality and depth percept. Analysis revealed that viewing experience and naturalness incorporated variations in image quality to a similar extent, yet the added value of stereoscopic depth is incorporated significantly more by naturalness. This result classifies naturalness as the most appropriate concept to evaluate 3D quality of stereoscopic stills. The 3D Quality Model based on naturalness as evaluation concept is validly applicable to stereoscopic stills and the naturalness score is determined for approximately 75% by image quality and for approximately 25% by the added value of stereoscopic depth.