Striate cortex extracts higher-order spatial correlations from visual textures.

Spatial correlations define the statistical structure of any visual image. Two-point correlations inform the visual system about the spatial frequency content of an image. Higher-order correlations can capture salient features such as object contours. We studied "isodipole" texture discrimination in V1 to determine if higher-order spatial correlations can be extracted by early stages of cortical processing. We made epicortical, local field potential, and single-cell recordings of responses elicited by isodipole texture interchange in anesthetized monkeys. Our studies demonstrate that single neurons in V1 can signal the presence of higher-order spatial correlations in visual textures. This places a computational mechanism, which may be essential for form vision at the earliest stage of cortical processing. IMAGES:

空间相关性定义了任意视觉图像的统计结构。两点相关性可向视觉系统传递图像的空间频率成分信息。高阶相关性则能够捕捉诸如物体轮廓这类显著视觉特征。我们在初级视皮层（V1）中开展等偶极子纹理辨别实验，以探究大脑皮层早期加工阶段能否提取高阶空间相关性。我们对麻醉猕猴实施记录，获取了等偶极子纹理交替刺激所引发的皮层表面电位、局部场电位以及单细胞响应。本研究证实，初级视皮层（V1）内的单个神经元能够表征视觉纹理中高阶空间相关性的存在。这表明在皮层加工的最早阶段，便存在一种可能对形态视觉至关重要的计算机制。图像：