Visual discomfort of 3D TV: Assessment methods and modeling

Abstract

To gain knowledge on how visual discomfort is built up while watching stereoscopic content an experiment was designed with two objectives: (1) to compare the continuous evaluation method with other assessment methods that potentially can substitute the continuous evaluation for the assessment of visual discomfort of e.g., feature-length movies, and (2) to relate the impact of time-variant content characteristics, such as motion and disparity, to the assessment of visual discomfort. In an experiment a 24 min 3D movie ‘Spy Kids 3-D: game over’ converted from 2D to 3D was displayed on a 9-view autostereoscopic lenticular LCD, and continuously assessed in terms of visual comfort by 24 participants. Additional assessment methods included the assessment of six 10 s sequences captured from the 3D movie and a single retrospective assessment of the entire 3D movie. Time-variant content characteristics, such as derivatives of motion and screen disparity values, were extracted from the 3D movie with motion and depth estimation algorithms. The moment-to-moment values of these characteristics were correlated to the continuous assessment scores of visual discomfort. With respect to the first objective, results reveal that the correlation between the assessment of the 10 s sequences captured from the 3D movie and their corresponding part within the continuous assessment is low, whereas the correlation between the retrospective assessment and the mean of the continuous assessment score over scene parts with a high screen disparity is higher. With respect to the second objective, for static scenes the visual comfort can be largely described by the screen disparity offset and range. For dynamic scenes the visual comfort is largely related to the screen disparity range, lateral motion and to the change in screen disparity.