Exploring mechanisms for the resolution of ambiguity in visual scenes

人类生活在一个不断变化的视觉环境中：视觉系统必须基于两个二维的视网膜输入，实时构建这个动态视觉世界的稳定的三维表示。视觉系统解决了实现无缝日常视觉体验的基本问题，使用复杂的底层计算和推理处理来简化和解决不可避免的歧义。这项工作着眼于其中的两种策略：视觉特征分组和视觉信息存储。双稳态刺激如内克尔立方体是视觉知觉重建的有用证明。立方体为视觉系统提供恒定的光学输入，但观察者的感知解释在两种选择之间切换。双稳定的另一个重要形式是双目竞争。当冲突的图像呈现在两只眼睛的对应点时（例如，一只眼睛的水平线，另一只眼睛的垂直线），感知在两个视图之间切换，而不是简单地合并它们。对物理稳定刺激的波动感知为研究神经处理提供了独特的机会。对竞争的传统解释涉及两只眼睛视觉信息之间的竞争，但最近的证据表明，竞争在等级上涉及一系列日益复杂的过程。例如，在模式竞争中，图像在两只眼睛之间被分割成碎片，但这可能会产生对整个图像的稳定感知。虽然这意味着机制比基于眼睛的竞争更复杂，但我们仍然不清楚不同层次的处理是如何协调的，或者当切换发生时是什么决定了视觉优先级。所提出的工作将通过关注对竞争的基于表面的影响来扩展当前对视觉处理层次的理解。我们生活在一个表面物体的世界里，视觉系统必须从视觉原语中吸收这些物体。因此，表面补全过程可能会影响双眼竞争目标的相对优势，这可以通过实验来检验。当考虑到我们的视觉体验时，所提出的研究的重要性是显而易见的：在观看双稳定刺激时，表现表面信息的优先级是否会影响我们的感知体验？回答这个问题将有助于我们深入了解感知系统是如何组织、简化和消除视觉输入歧义的。视觉信息的存储视觉系统将来自许多不同来源（双目线索、阴影、透视等）的关于物体或场景的信息组合在一起。然而，当物体在视觉上被遮挡时，或者当照明（或其他因素）的变化改变了进入眼睛的光线模式时，视觉输入可能会中断或退化，因此能够随着时间的推移整合视觉信息以获得有关物体属性的可靠信息是有利的。众所周知，视觉信息可以被“储存”，并在以后的时间里用于感知决策，但其中的机制尚不清楚。这是第二组实验的重点，它将使用成像技术与心理物理学相结合来阐明感知决策的机制。对相关视觉属性（运动、视差）做出反应的大脑区域是通过保持其活动来存储信息，还是存在单独的区域来整合跨属性的信息？储存的时间是怎样的？保存的时间是多长？强调负责视觉信息的时间整合和存储的大脑机制将大大增加我们对视觉功能的理解。

Humans live in a constantly changing visual environment: The visual system must construct a stable, three-dimensional representation of this dynamic visual world in real time, based primarily on two, two-dimensional retinal inputs. The visual system solves the fundamental problem of achieving seamless everyday visual experience using complex underlying computational and inferential processing to simplify and resolve inevitable ambiguities. The proposed work looks at two of these strategies: grouping of visual features and storage of visual information. Grouping of visual features Bi-stable stimuli such as the Necker Cube are useful demonstrations of vision as perceptual reconstruction. The Cube provides a constant optical input to the visual system but the observer's perceptual interpretation switches between two alternatives. Another important form of bi-stability is binocular rivalry. When conflicting images are presented to corresponding points in the two eyes (e.g. horizontal lines to one eye, vertical to the other), perception switches between the two views, rather than simply merging them. Fluctuating perceptions of physically stable stimuli offer unique opportunities to study neural processing. Traditional accounts of rivalry involve a competition between the two eyes' visual information but recent evidence has demonstrated that rivalry hierarchically engages a set of increasingly elaborate processes. For example, in pattern rivalry, an image is divided piecemeal between the two eyes but this may produce stable perception of the entire image. Although this implies mechanisms that are more complex than eye-based competition, it remains unclear how the different levels of processing are coordinated or what determines visual priority when switching occurs. The proposed work will extend current understanding of the visual processing hierarchy by focusing on surface-based influences on rivalry. We exist in a world of surface-based objects, which the visual system must assimilate from visual primitives. Thus, surface completion processes may influence the relative dominance of binocular rivalry targets and this can be examined experimentally. The importance of the proposed research is apparent when considering our visual experience: Does the priority for representing surface information influence our perceptual experience when viewing bi-stable stimuli? Answering this question would provide a useful insight into how the perceptual system organises, simplifies, and disambiguates visual input. Storage of visual information The visual system combines information about an object or scene from many different sources (binocular cues, shading, perspective, etc.). Because visual input may, however, be disrupted or degraded when objects are obscured from vision, or when changes in lighting (or other factors) alter the pattern of light entering the eye, it is advantageous to be able to integrate visual information over time to obtain reliable information regarding object properties. It is known that visual information can be 'stored' and used to inform perceptual decisions at a later time, but the mechanisms involved are as yet unclear. This is the focus of the second set of experiments, which will use imaging techniques in combination with psychophysics to elucidate the mechanisms of perceptual decision making. Do brain areas responding to relevant visual attributes (motion, disparity) store information by maintaining their activity or do separate regions exist that integrate information across attributes? What is the time course of storage and how long is it maintained? Highlighting brain mechanisms responsible for the temporal integration and storage of visual information would add substantially to our understanding of visual functioning.