Visual Perception of Motion and Texture

本研究旨在了解人类视觉感知的长期目标，特别是在日常生活中与动物和人类功能相关的低水平机制。这项特别的工作集中在视觉世界的基本方面，如亮度、对比度、纹理、运动和颜色，是如何被人类大脑用来从背景中看到物体的（图-地分割），并判断不同距离上物体的相对深度。这些项目将回答以下问题：(1)高阶图像统计（如稀疏性、局部边缘结构）在纹理分割中的作用；(2)亮度和颜色在纹理分割中的作用；(3)运动视差对表面纹理的影响；(4)运动视差时不完全凝视行为的影响。这些目标将通过定量心理物理实验和眼动记录来实现，并通过心理物理数据测试数学模型。新的方法方面包括：(1)使用现代计算机图形渲染对运动视差进行定量研究，(2)从自然图像或我们最近开发的模拟自然图像统计的合成纹理构建视觉刺激，以及(3)运动视差心理物理学期间凝视方向的详细分析。图-地分割是视觉研究中的一个关键问题，它超越了我们目前对早期视觉处理的认识，是随后的物体识别、空间认知和视觉运动控制的必要前提，到目前为止，它已经使用高度人工刺激进行了研究。这项研究主要是为了提高我们对视觉感知的基本理解，特别是在更自然的情况下，例如运动视差和复杂的纹理。从长远来看，研究结果可能有助于计算机视觉、虚拟现实显示和临床诊断刺激设计的创新，以评估视力障碍。

This research is aimed at the long-term objective of understanding human visual perception, particularly low-level mechanisms that are functionally relevant to animals and people in everyday life. This particular work is focused on how elementary aspects of the visual world, such as luminance, contrast, texture, motion and colour, are used by the human brain to see objects stand out from their backgrounds (figure-ground segmentation), and to judge relative depths of objects at different distances. The projects will answer questions about (1) the role of higher-order image statistics (e.g. sparseness, local edge structure) in texture segmentation; (2) the role of luminance and colour in texture segmentation; (3) how motion parallax performance depends on surface texture; and (4) the implications of imperfect gaze behaviour during motion parallax. These goals will be achieved using quantitative psychophysical experiments and eye movement recording, with mathematical models which are tested on psychophysical data. Novel methodological aspects include (1) the quantitative study of motion parallax using modern computer graphics rendering, (2) the construction of visual stimuli from natural images or our recently developed synthetic textures designed to mimic the statistics of natural images, and (3) detailed analysis of gaze direction during motion parallax psychophysics. Figure-ground segmentation is a key problem in vision research, going beyond our present knowledge of early visual processing and a necessary precursor to subsequent object recognition, spatial cognition, and visuomotor control - until now it has been studied using highly artificial stimuli. This research is primarily directed towards improving our basic understanding of visual perception, particularly in the context of more naturalistic situations, e.g. motion parallax and complex textures. In the longer term, the results may contribute to innovations in computer vision, virtual reality displays, and the design of clinical diagnostic stimuli for assessing visual impairment.