立体镜视觉研究的回顾、反思与展望  

Review,Introspection and Prospects on the Stereoscopic Vision Studies

作　　者：王天珍[1] 

机构地区：[1]武汉理工大学自动化学院,湖北武汉430070

出　　处：《武汉理工大学学报（信息与管理工程版）》2011年第1期1-5,共5页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家自然科学基金资助项目(60275040)

摘　　要：基于对立体镜视觉研究史中重大事件的分析,发现近20多年来引起主流理论重大危机的原因可归纳为4点:将错觉当成正常知觉来研究,过分夸大了视差在深度知觉中的作用;将等待视系统输出决定的视差当作描述输入最主要的量来计算深度,使计算视觉的问题成为病态问题;视差是一个局域量,而影响深度加工的双眼差别是全局性的,因此用视差描述双眼视觉信息加工过程是不合适的;3D重建理论属于先天论,没有考虑经验和眼运动的影响。从整体研究双眼视觉是今后研究的方向,即不只考虑整个网膜图像,还意味着未来的双眼深度模型不仅能解释立体视觉,也能解释其他双眼现象,如竞争、复视和叠加等。Based on the analyses of the important events in 20 years history of Stereoscopic Vision studies,the reasons which induced these serious crises in prevalent stereopsis theory can be summed up as the following 4 viewpoints.In the studies,illusion was taken as normal visual perception,thus the role of disparity in the depth perception was exaggerated.Being an output of visual system,disparity was used as an input in 3D reconstruction computation theory,so it made vision computation an ill-posed problem.As a local variable,disparity was not suitable to describe the global binocular information processing that was related to difference of both eyes images.3D reconstruction computation theory holds up nativism in visual science without taking visual experience and ocular motion into consideration.In view of the future trend,the binocular vision research should be conducted in a global way.Here the global way means that the researches should be focused on the whole retina image,and the new model can explain not only stereopsis,but also other binocular phenomena,such as rivalry,diplopia,superposition,etc.

关 键 词：立体镜视觉 错觉 视差 3D重建理论 双眼视觉的视动反馈模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]