Tsirlin Inna, Wilcox Laurie M, Allison Robert S
Centre for Vision Research, York University, Toronto, ON, CanadaEye Movement and Vision Neuroscience Laboratory, The Hospital for Sick Children, Toronto, ON, Canada.
Centre for Vision Research, York University, Toronto, ON, Canada.
J Vis. 2014 Jun 9;14(7):5. doi: 10.1167/14.7.5.
In binocular vision, occlusion of one object by another gives rise to monocular occlusions—regions visible only in one eye. Although binocular disparities cannot be computed for these regions, monocular occlusions can be precisely localized in depth and can induce the perception of illusory occluding surfaces. The phenomenon of depth perception from monocular occlusions, known as da Vinci stereopsis, is intriguing, but its mechanisms are not well understood. We first propose a theory of the mechanisms underlying da Vinci stereopsis that is based on the psychophysical and computational literature on monocular occlusions. It postulates, among other principles, that monocular areas are detected explicitly, and depth from occlusions is calculated based on constraints imposed by occlusion geometry. Next, we describe a biologically inspired computational model based on this theory that successfully reconstructs depth in a large range of stimuli and produces results similar to those described in the psychophysical literature. These results demonstrate that the proposed neural architecture could underpin da Vinci stereopsis and other stereoscopic percepts.
在双眼视觉中，一个物体被另一个物体遮挡会产生单眼遮挡——即仅在一只眼睛中可见的区域。尽管无法为这些区域计算双眼视差，但单眼遮挡在深度上可以被精确地定位，并且可以诱发对虚幻遮挡表面的感知。从单眼遮挡产生深度感知的现象，即所谓的达·芬奇立体视觉，很有趣，但其机制尚未得到很好的理解。我们首先基于关于单眼遮挡的心理物理学和计算文献，提出一种关于达·芬奇立体视觉潜在机制的理论。该理论除其他原则外，假定单眼区域被明确检测到，并且基于遮挡几何形状所施加的约束来计算遮挡产生的深度。接下来，我们描述一个基于该理论的受生物启发的计算模型，该模型成功地在大范围的刺激中重建深度，并产生与心理物理学文献中描述的结果相似的结果。这些结果表明，所提出的神经架构可能是达·芬奇立体视觉和其他立体视觉感知的基础。
