基于视差和串扰的自由立体显示器视疲劳预测

Prediction of visual fatigue for autostereoscopic displays based on parallax and crosstalk

针对观看自由立体显示器存在人眼焦点调节和集合不一致的根本生理问题,及观看时间稍长将引起立体观看视疲劳,甚至危害视觉健康的问题。提出了一种对自由立体显示器视疲劳程度进行预测的方法,由三个不同的立体观看视疲劳参数--视差值、视频帧率和串扰度组合得到27段立体测试视频,让观看者观看这些视频得到主观评价数据,采用多元线性回归的统计方法处理评价数据确定了自由立体显示器视疲劳的多元线性回归模型,并根据此模型对待评估立体视频的视疲劳程度进行有效预测。这种方法避免了反复的主客观评价对视觉系统的危害,减少主客观评价信息收集的复杂性。最重要的是,该模型能快速预测自由立体显示器视疲劳程度。

One fundamental problem arising from viewing contents in autostereoscopic displays is the conflict between accommodation and convergence. Prolonged viewing may result in symptoms of severe visual fatigue, producing harmful effect on the visual health. A method is proposed to predict the degree of visual fatigue for autostereoscopic displays.Subjects view 27 stereoscopic video which are sequences generated by software. The stereoscopic sequences differ in three factors including parallax magnitude, frame rate, and crosstalk, which are supposed to exert substantial influence upon visual fatigue. Combined with subjective evaluation method, the linear regression of mathematical statistics is used to deal with the observation data. The multivariate linear regression model of visual fatigue for autostereoscopic displays is determined. Based on the model, the degree of visual fatigue caused by stereoscopic video could be estimated effectively. According to the experimental results, this model prevent damage to human vision system resulting from reduplicative process of subjective and objective evaluation. It also reduces the complexity of information collection. And above all, the proposed model can quickly predict the degree of visual fatigue for autostereoscopic displays.