Recovery of Viewer-Centered and Object-Centered Depth from 2-D Projections

在投影到两只眼睛视网膜上的二维图像的基础上感知三维环境需要整合不同来源的深度信息。该项目将研究两种深度信息的整合，即以观察者为中心的信息，它提供观察者与环境中物体之间的相对距离，以及以物体为中心的信息，它提供外部场景内的相对距离，与观察者的位置无关。将被研究的以观察者为中心的信息将包括双眼视差（物体投射到两只眼睛视网膜上的差异，作为与观察者的相对距离的函数）和运动视差（投射到眼睛上的速度变化，作为与观察者的距离的函数）。以物体为中心的信息主要是运动结构（在刚性运动的假设下，使用改变投影中的距离来恢复三维物体上特征之间的距离）。一系列的实验将研究如何通过将视网膜图像中的运动划分为以观察者为中心和以物体为中心的组件（运动视差和运动结构）来确定物体的感知三维形状。第二个系列的实验将检验从以物体为中心的信息中恢复的形状整合到以观察者为中心的三维环境表示中。第三个系列的实验将检验当在三维场景中添加指示缺乏深度的信息时，物体的感知形状和方向的变化。第四个系列的实验将确定从以观察者为中心的信息（双眼视差）和以物体为中心的信息（结构-运动）的组合中检测表面所需的物体上可见特征的最小数量。本研究的目的是了解如何将不同类型的深度信息结合起来，以提供对三维环境的综合感知。这与在实际三维环境中优化人类在物体识别方面的表现以及设计人工三维显示器（包括飞行模拟器的显示器和代表虚拟环境的飞机显示器）等问题相关。

Perceiving the three-dimensional environment on the basis of the two-dimensional images projected onto the retinas of the two eyes requires integration of different sources of depth information. This project will examine the integration of two types of depth information, i.e., viewer-centered information, which provides relative distances between an observer and objects in the environment, and object-centered information, which provides relative distances within an external scene, independent of viewer location. The viewer-centered information that will be studied will include binocular disparity (differences in the object's projection onto the retinas of the two eyes as a function of relative distances from the observer) and motion parallax (variations in the velocities projected onto the eye as a function of distance from the observer). The object-centered information of primary interest is structure-from-motion (use of changing distances in the projection to recover distances between features on a three-dimensional object, under an assumption of rigid motion). One series of experiments will examine how the perceived three-dimensional shapes of objects are determined by the partitioning of motion in the retinal images into viewer-centered and object-centered components (motion parallax and structure-from- motion). A second series of experiments will examine integration of shape recovered from object-centered information into a viewer- centered representation of the three-dimensional environment. A third series of experiments will examine the changes in the perceived shapes and orientations of objects that occurs when information indicating lack of depth is added to a three- dimensional scene. A fourth series of experiments will determine the minimum number of visible features on an object needed to detect a surface from a combination of viewer-centered information (binocular disparity) and object-centered information (structure- from-motion). The objective of this research is to understand how different types of depth information are combined to provide an integrated perception of the three-dimensional environment. This has relevance for such issues as optimizing human performance in object recognition in actual three-dimensional environments and designing artificial three-dimensional displays, including displays for flight simulators and aircraft displays that represent the virtual environment.