Kovács I, Julesz B
Laboratory of Vision Research, Rutgers University, Piscataway, New Jersey 08854.
Nature. 1994 Aug 25;370(6491):644-6. doi: 10.1038/370644a0.
An unsolved problem of biology is the processing of global shape in natural vision. The known processes of early vision are spatially restricted (or local) operations, and little is known about their interactions in organizing the visual image into functionally coherent (or global) objects. Here we introduce a human psychophysical method which allows us to measure the effect of perceptual organization on the activity pattern of local visual detectors. We map differential contrast sensitivity for a target across regions enclosed by a boundary. We show that local contrast sensitivity is enhanced within the boundary even for large distances between the boundary and the target. Furthermore, the locations of maximal sensitivity enhancement in the sensitivity maps are determined by global shape properties. Our data support a class of models which describe shapes by the means of a medial axis transformation, implying that the visual system extracts 'skeletons' as an intermediate-level representation of objects. The skeletal representation offers a structurally simplified shape description which can be used for higher-level operations and for coding into memory.
生物学中一个尚未解决的问题是自然视觉中全局形状的处理。已知的早期视觉过程是空间受限（或局部）的操作，而对于它们在将视觉图像组织成功能连贯（或全局）对象时的相互作用却知之甚少。在这里，我们介绍一种人类心理物理学方法，该方法使我们能够测量知觉组织对局部视觉探测器活动模式的影响。我们绘制了边界所包围区域内目标的差分对比敏感度。我们表明，即使边界与目标之间距离很大，边界内的局部对比敏感度也会增强。此外，敏感度图中敏感度增强最大值的位置由全局形状属性决定。我们的数据支持一类通过中轴变换来描述形状的模型，这意味着视觉系统提取“骨架”作为对象的中级表示。骨架表示提供了一种结构简化的形状描述，可用于更高层次的操作和编码到记忆中。
