Stereopsis in natural environments

人类视觉系统最基本的作用之一是提供我们环境中物体的深度和距离信息，以便我们能够安全有效地与我们的世界互动。从拿咖啡这样简单的日常目标，到进行显微手术所需的精密程度，深度判断都是必不可少的。在个人互动空间中，双眼深度提示立体视觉是我们获得相对深度信息最精确、最准确的来源。然而，关于仅从立体视觉获得3D深度感知的准确性，以及与其他深度线索的结合，存在相当大的争议。我认为，这一争议的主要原因是，在大多数实验中，由于使用3D显示器和简化的渲染刺激而导致的深度线索存在冲突。这一系列实验的目的是通过量化立体视觉和单眼线索在使用真实物体感知深度时的相互作用来解决这一争论。为了做到这一点，我将使用一种独一无二的计算机控制设备，通过控制灯光和运动，在精确的位置呈现物理对象。刺激将使用先进的建模软件和3D打印精确的规格设计；我们将系统地操纵深度线索的存在和一致性，如双目视差，纹理梯度，焦点模糊和运动视差。所有实验都将使用心理物理方法（例如恒定刺激方法），并使用概率模型来比较深度线索与理论预测的组合。该研究项目将在以下方面产生重大影响：1)弥合使用计算机生成刺激的研究与我们如何在自然世界中使用深度信息之间的差距；2)解决长期以来关于个体深度线索度量性质的争论；3)阐明立体视觉在自然条件下深度感知中的作用。这里提出的实验是一个更大的研究项目的核心，最终将定义立体深度感知在我们与世界有效互动的能力中的影响和重要性。研究结果为自然环境下视觉线索组合的研究提供了重要的参考依据。这项工作的意义是深远的，因为立体视觉是从娱乐到显微外科和飞行模拟等广泛领域的3D可视化的基石。

One of the most fundamental roles of the human visual system is to provide information about the depth and distance of objects in our environment so that we can interact with our world safely and effectively. Depth judgments are essential for tasks ranging from simple day-to-day objectives like reaching for your coffee, to the exquisite precision needed to perform microsurgery. Within personal interaction space, the binocular depth cue stereopsis is our most precise and accurate source of relative depth information. However, there is considerable debate concerning the accuracy of 3D depth percepts from stereopsis alone, and in combination with other depth cues. I propose that a primary cause of this controversy is the presence of conflicting depth cues in most experiments caused by the use of 3D displays and simplified rendered stimuli. The aim of this series of experiments is to resolve this debate by quantifying the interaction between stereopsis and monocular cues in the perception of depth using real physical objects. To do so, I will use a one-of-a-kind computer-controlled apparatus to present physical objects at precise locations with controlled lighting and motion. Stimuli will be designed using advanced modelling software and 3D printed to exact specifications; we will systematically manipulate the presence and consistency of depth cues such as binocular disparity, texture gradients, focal blur and motion parallax.  All experiments will use psychophysical methods (e.g. the method of constant stimuli) and probabilistic models will be used to compare the combination of depth cues with theoretical predictions. This research program will have significant impact in: 1) bridging the gap between studies using computer generated stimuli, and how we use depth information in the natural world 2) addressing the long-standing debate concerning the metric nature of individual depth cues and 3) clarifying the role of stereopsis in depth perception under natural conditions. The experiments proposed here are the core of a larger research program that ultimately will define the impact and importance of stereoscopic depth perception in our ability to effectively interact with the world. The results provide an important benchmark for studies of visual cue combination in natural environments. The implications of the work are far-reaching, for stereopsis is the cornerstone of 3D visualization in a wide range of fields from entertainment to microsurgery and flight simulation.