Lunn P D, Morgan M J
Institute of Ophthalmology, University College, London, UK.
J Opt Soc Am A Opt Image Sci Vis. 1997 Feb;14(2):360-71. doi: 10.1364/josaa.14.000360.
Observers discriminated the relative disparity, disparity gradient, and disparity curvature of surfaces defined by horizontal binocular disparity in random-dot stereograms. In experiment 1, thresholds for discriminating the depth of sinusoidal corrugations were very similar for different corrugation frequencies, despite large differences in disparity gradient and disparity curvature. Thus observers used a relative disparity cue in preference to a slant or curvature cue. Experiment 2 isolated the spatial derivatives of disparity by jittering the other available cues, using surfaces with square-wave, triangle-wave, and parabolic-wave profiles. Weber fractions were 4%-10% for relative disparity, 6%-12% for disparity gradient, and 15%-30% for disparity curvature. Experiment 3 confirmed this result for larger surfaces. The study supports the view that human stereoscopic vision aims to represent the local scene relative to the observer, at the expense of computing intrinsic properties of objects, such as curvature.
观察者能够辨别由随机点立体图中水平双眼视差所定义的表面的相对视差、视差梯度和视差曲率。在实验1中,尽管视差梯度和视差曲率存在很大差异,但对于不同波纹频率的正弦波纹深度辨别阈值非常相似。因此,观察者优先使用相对视差线索而非倾斜或曲率线索。实验2通过抖动其他可用线索,使用具有方波、三角波和抛物线波轮廓的表面,分离了视差的空间导数。相对视差的韦伯分数为4%-10%,视差梯度为6%-12%,视差曲率为15%-30%。实验3对更大的表面证实了这一结果。该研究支持这样一种观点,即人类立体视觉旨在相对于观察者呈现局部场景,而牺牲了计算物体的固有属性,如曲率。