Barenholtz Elan, Feldman Jacob
Brown University, Department of Cognitive and Linguistic Sciences, 190 Thayer St., Providence, RI 02912, USA.
Cognition. 2006 Oct;101(3):530-44. doi: 10.1016/j.cognition.2005.12.002. Epub 2006 Jan 31.
Figure/ground assignment - determining which part of the visual image is foreground and which background - is a critical step in early visual analysis, upon which much later processing depends. Previous research on the assignment of figure and ground to opposing sides of a contour has almost exclusively involved static geometric factors - such as convexity, symmetry, and size - in non-moving images. Here, we introduce a new class of cue to figural assignment based on the motion of dynamically deforming contours. Subjects viewing an animated, deforming shape tended to assign figure and ground so that articulating curvature extrema - i.e., "hinging" vertices - had negative (concave) contour curvature. This articulating-concavity bias is present when all known static cues to figure/ground are absent or neutral in each of the individual frames of the animation, and even seems to override a number of well-known static cues when they are in opposition to the motion cue. We propose that the phenomenon reflects the visual system's inbuilt expectations about the way shapes will deform - specifically, that deformations tend to involve rigid parts articulating at concavities.
图形/背景分配——确定视觉图像的哪一部分是前景，哪一部分是背景——是早期视觉分析中的关键步骤，许多后续处理都依赖于此。先前关于将图形和背景分配到轮廓相对两侧的研究几乎完全涉及静态几何因素——如凸性、对称性和大小——在非移动图像中。在这里，我们基于动态变形轮廓的运动引入了一种新的图形分配线索类别。观看动画变形形状的受试者倾向于分配图形和背景，以使关节曲率极值——即“铰接”顶点——具有负（凹）轮廓曲率。当动画的每个单独帧中所有已知的图形/背景静态线索都不存在或呈中性时，这种关节凹度偏差就会出现，甚至当一些众所周知的静态线索与运动线索相反时，它似乎会凌驾于这些线索之上。我们提出，这种现象反映了视觉系统对形状变形方式的内在期望——具体来说，变形往往涉及在凹处铰接的刚性部分。
