Subjective contours capture stereopsis
Subjective contours capture stereopsis
复制标题
主观轮廓捕捉立体视觉
作者：
V. Ramachandran;P. Cavanagh
立体深度知觉是基于测量两只眼睛图像之间的微小差异，这些差异是双目视差1的结果。Julesz2使用随机点立体图证明，立体视觉可以基于两只眼睛图像的简单点对点比较，而不需要单眼可见的形状或轮廓的存在。在这里，我们提出了一类新的立体图，它表明单目主观轮廓3-5可以影响立体图中元素的匹配，即使元素本身不传递视差信息。更具体地说，从这些轮廓中看到的深度自动归因于这些轮廓所包围的纹理元素和线条--这是一种我们称为“立体捕捉”的错觉。
Stereoscopic depth perception is based on measuring tiny differences between the two eyes' images which arise as a result of binocular parallax1. Julesz2 used random-dot stereograms to demonstrate that stereopsis may be based on a simple point-to-point comparison of the two eyes' images and does not require the presence of monocularly visible forms or contours. Here we present a new class of stereograms which illustrate that monocular subjective contours3–5 can influence the matching of elements in a stereogram even though the elements themselves convey no disparity information. More specifically, the depth seen from such contours is automatically attributed to texture elements and lines that are enclosed by these contours—an illusion that we call ‘stereoscopic capture’.