MECHANISMS OF STEREOSCOPIC DEPTH RECONSTRUCTION

描述(改编自《调查者摘要》)：目标是 从三个层面分析加工机制的空间结构 人类深度重建过程的一部分。这项提议的动机是 用于将双目信息汇聚成一个 观察者面对的3D表面的统一深度图。尽管过去 关于深度处理的工作，深度重建的许多性质 这一框架每一级的进程仍未解决，特别是 在每个处理级别上的频道结构的调谐。这项建议 重点评价了三个层次的框架与三个 心理物理工具：深度的可探测性，深度的后效 视差和单眼深度线索的适应及其形式后效 在没有深度适配的情况下对深度图像的形式进行适配。 这些工具将被应用于确定 深度加工的输入水平，深度的机制 重建和识别得到的深度图像的形式。 尽管立体视觉通常被视为 感知深度，还有许多其他单眼深度线索可以 并行处理以生成戏剧性的深度印象(例如 不同的运动提示、透视、明暗处理形状、纹理渐变、 模糊、插补等)，其与立体深度的交互 CUES最近引起了人们的极大兴趣。该提案的一个关键主题是 将这些机制融合到3D的统一映射中的程度 与观察者对峙的表面。

DESCRIPTION (Adapted from the Investigator's Abstract): The goal is to analyze the spatial structure of the processing mechanisms at three levels of the human depth reconstruction process. The proposal is motivated by a 5-stage framework for the convergence of binocular information into a unified depth map of the 3D surfaces confronting the observer. Despite past work on depth processing, many properties of the depth reconstruction process at each level of this framework are still unresolved, particularly the tunings of the channel structure at each processing level. The proposal focuses on evaluation of three levels of the framework with three psychophysical tools: detectability of depth, the depth aftereffect of adaptation to disparity and monocular depth cues, and the form aftereffect of adaptation to the form of a depth image without depth adaptation as such. These tools will be applied to determine the functional organization at the level of the inputs to depth processing, the mechanisms of depth reconstruction and the recognition of the form of the resulting depth image. Although stereopsis is generally viewed as the strongest component of perceived depth, there are many other monocular depth cues that can be processed in parallel to generate dramatic depth impressions (such as differential motion cues, perspective, shape-from-shading, texture gradient, blur, interposition and so on), whose interaction with stereoscopic depth cues has generated much recent interest. A key topic of the proposal is the degree of convergence of these mechanisms into a unified map of the 3D surfaces confronting the observer.