Neural Basis of Shape from Texture

描述（由申请人提供）：本项目的目的是了解图案表面的三维形状感知。外纹状皮层包括许多神经元，这些神经元对由纹理线索定义的3-D表面方向具有选择性，但是没有建立这种感受野的神经连接的可行模型。我们已经表明，从纹理中提取3-D形状，纹理均匀性的假设和纹理梯度的估计是不需要的。相反，表面曲率/取向估计由取向流提供，深度估计由频率流提供。这些信息流预测正确的感知和错误的感知。由于V1神经元对方向和频率有选择性，我们将建立模型，联合收割机结合它们的输出来提取方向和频率流，以识别不同的形状特征。实验将详细说明神经机制的时间和空间特性。我们将使用新类型的诊断刺激，没有线索冲突，检查的方式，其中运动和纹理线索在3-D感知中相互作用，以及运动是否有助于感知的内在曲率或只是相对深度。我们将在方向流和频率梯度方面分析自然纹理物体上的阴影图案。我们将看到阴影的这些方面如何在3D感知中与纹理的方向和频率信息相互作用。我们将研究纹理和轮廓线索如何相互作用，以解释真实的3-D固体的一些运动和形状错觉。我们将测试神经机制的形状从纹理是硬连线或可以改变的经验，通过使用触觉反馈，以纠正误解的纹理线索。这项研究将展示简单的神经连接如何导致复杂的三维感知。这些结果将有助于了解形状感知神经缺陷的原因，并为有这种缺陷的患者建立人工视觉辅助工具。

DESCRIPTION (provided by applicant): The aim of this project is to understand the three-dimensional shape perception of patterned surfaces. Extra-striate cortex includes many neurons that are selective to 3-D surface orientation defined by texture cues, but there are no feasible models of neural connections that build such receptive fields. We have shown that to extract 3-D shape from texture, assumptions of texture homogeneity and estimates of texture gradients are not needed. Instead, surface curvature/orientation estimates are provided by orientation flows, and depth estimates by frequency flows. These flows predict both correct perceptions and misperceptions. Since V1 neurons are selective for orientation and frequency, we will build models that combine their outputs to extract orientation and frequency flows for identifying distinct shape features. Experiments will specify the temporal and spatial properties of the neural mechanisms. We will use new types of diagnostic stimuli, without cue conflict, to examine the manner in which motion and texture cues interact in 3-D perception, and whether motion contributes to the perception of intrinsic curvature or just to relative depth. We will analyze shading patterns on naturalistic textured objects in terms of orientation flows and frequency gradients. We will see how these aspects of shading interact in 3-D perception, with orientation and frequency information from textures. We will examine how texture and silhouette cues interact to explain some motion and shape illusions with real 3-D solids. We will test whether neural mechanisms for shape-from texture are hard-wired or can be altered by experience, by using haptic feedback to correct misperceptions from texture cues. This study will show how simple neural connections lead to complex 3-D percepts. These results will be useful both in understanding the causes of neurological deficits in shape perception, and in building artificial visual aids for patients with such deficits.