Abstract Geometric models are tools used to analyze geometric distortions in S3D viewing. For standard geometric models, 3D perception is predicted by calculating the intersection between the two projection lines from the left and the right eye to two corresponding onscreen points. However, standard geometric models fail in two conditions: (1) S3D images are captured by converging cameras (toe-in) and displayed on a flat screen; (2) the viewer’s head is rotated about the vertical axis (yaw) or depth axis (roll) relative to the screen. In these two conditions, vertical screen disparity with respect to the viewer is introduced. In this paper, we extend standard geometric models to cover the conditions with vertical screen disparity. To test our models, we conduct experiments using random-dot stereograms (RDS) with vertical screen disparities introduced by rotating RDS images. Experimental results show that the perceived depth is well predicted by the proposed model when the rotation angle is smaller than 20 ° . However, this paper introduces a retinal eccentricity model to analyze geometric distortions from the viewer’s perspective, which indicates that the geometric distortions only affect the binocular depth cue and not monocular depth cues. The monocular-binocular depth cue conflicts with motions may induce motion sickness in S3D viewing.

中文翻译：

---

具有垂直屏幕视差的立体 3D 的扩展几何模型

摘要 几何模型是用于分析 S3D 视图中几何失真的工具。对于标准几何模型，通过计算从左眼和右眼到两个对应屏幕点的两条投影线之间的交点来预测 3D 感知。然而，标准几何模型在两种情况下会失败：（1）S3D 图像由会聚相机（前束）捕获并显示在平面屏幕上；(2) 观看者的头部相对于屏幕绕垂直轴（yaw）或深度轴（roll）旋转。在这两种情况下，引入了相对于观看者的垂直屏幕差异。在本文中，我们扩展了标准几何模型以涵盖具有垂直屏幕视差的条件。为了测试我们的模型，我们使用随机点立体图 (RDS) 进行实验，通过旋转 RDS 图像引入垂直屏幕视差。实验结果表明，当旋转角度小于20°时，所提出的模型能够很好地预测感知深度。然而，本文引入了一个视网膜偏心模型来从观察者的角度分析几何畸变，这表明几何畸变只影响双眼深度线索，而不影响单眼深度线索。单眼-双眼深度线索与运动的冲突可能会导致 S3D 观看中的晕车。本文介绍了一种视网膜偏心模型，从观察者的角度分析几何畸变，表明几何畸变只影响双眼深度线索，而不影响单眼深度线索。单眼-双眼深度线索与运动的冲突可能会导致 S3D 观看中的晕车。本文介绍了一种视网膜偏心模型，从观察者的角度分析几何畸变，表明几何畸变只影响双眼深度线索，而不影响单眼深度线索。单眼-双眼深度线索与运动的冲突可能会导致 S3D 观看中的晕车。