Anderson B L, Nakayama K
Department of Psychology, Harvard University.
Psychol Rev. 1994 Jul;101(3):414-45. doi: 10.1037/0033-295x.101.3.414.
Models of stereopsis have focused on developing strategies for identifying common features in the 2 half-images so that disparity may be computed. This emphasis ignores the unpairable features that arise at occluding contours (half-occlusions). Most models treat half-occlusions as noise or outliers that are interpreted after disparity processing is completed. A series of experiments reveal that occlusion relationships are sensed during the earliest stages of binocular processing. The authors hypothesize the existence of receptive field structures that sense the local structure of stereoscopic occlusion relationships to account for these findings. Finally, a simple theoretical framework is presented in which fusion, stereopsis, and occlusion are unified. This theory explains the co-occurrence of stereopsis and diplopia and how half-occlusions escape the suppression characteristic of binocular rivalry.
立体视觉模型一直专注于开发识别两个半图像中共同特征的策略,以便能够计算视差。这种重点忽略了在遮挡轮廓(半遮挡)处出现的无法配对的特征。大多数模型将半遮挡视为噪声或异常值,在视差处理完成后再进行解释。一系列实验表明,在双眼处理的最早阶段就能感知到遮挡关系。作者推测存在能够感知立体遮挡关系局部结构的感受野结构来解释这些发现。最后,提出了一个简单的理论框架,将融合、立体视觉和遮挡统一起来。该理论解释了立体视觉和复视的同时出现,以及半遮挡如何避开双眼竞争的抑制特性。