Abstract
Stereopsis is a powerful mechanism for 3D object perception. However, in the absence of sufficient surface texture the stereoscopic information is often ambiguous. For instance, the disparity of a vertical contrast border provides depth information, but does not indicate whether the border corresponds to an edge formed by two surfaces, as in a roof, or an occluding edge, and whether the occluding edge belongs to the surface on the left or the surface on the right. We used single cell recording from awake fixating monkeys to explore how the visual cortex combines monocular form cues with stereo information.
(1) Many cells signal border ownership for figures in uniform surface displays, which requires processing of monocular form cues (Zhou et al., 2000). We found that about 1/3 of these cells combine monocular border ownership cues with selectivity for edges in random-dot stereograms and differentiate step edge polarity. They were usually tuned to some disparity D1 for one side of the edge, and responded if and only if the disparity on the other side is “farther” than D1. We found that the “near” side generally agreed with the preferred side of ownership. (2) Many cells differentiated between convex and concave roof edges in uniform surface displays. Thus, a contrast border was represented as a 3D feature on the basis of the disparities of distant contours. (3) Some cells showed interaction between roof shape and edge contrast polarity. In 10 out of 11 the interaction was in the same direction as in human vision which implies illumination from above. These signals contribute shape from shading. (4) Cells that were convex-concave selective with random-dot stereograms were generally not selective with uniform surface displays, and vice versa.
These results suggest that stereoscopic and contextual monocular form cues are combined in cells of monkey area V2. There seem to be several stages in this process.
Supported by NIH NS38034 and NIH EY02966