Configuration Effects in the Stereoprocessing of 3D Surfaces

Document Type


Publication Date






Publication Title

Journal of Vision

Subject Categories



Recently (ARVO 2001) we showed that stereoacuity in 3D surfaces was affected by shape(flat vs. curved) and by the spatial distribution of disparity signals across the stimulus surface (disparate contours vs. random-dot stereograms). For equivalent amounts of disparity, curved uniform-luminance surfaces with disparate contours had much higher thresholds than flat uniform-luminance surfaces or curved random-dot surfaces. One explanation for this result is that spatially continuous disparity signals cause disparity detectors to interact. If an interaction does occur, it should be possible to measure its vertical and horizontal spatial extent. To this end, flat and curved 3D surfaces (2.3o sq.) were generated using vertically-spaced horizontal lines. Flat surfaces were depicted with up to 19.6′ of standing disparity to equal the disparity of the most curved surface. Disparity thresholds were measured using observers' judgments of whether a line probe with variable disparity was in front of or behind the center of each surface. The vertical extent of interaction was estimated by varying the vertical spacing between the lines. The horizontal extent of interaction was estimated by replacing solid lines with segmented lines. The duty cycle was fixed at 50% while the segment length varied. Thresholds for surfaces with large curvature disparity were a factor of 2 higher than those for flat surfaces with equivalent standing disparity when lines were vertically separated by .5 deg. Thresholds for curved surfaces decreased dramatically, however, when line segments were horizontally separated by ∼ .5 deg. These thresholds were indistinguishable from those for a random dot stereogram with the same overall luminance, ruling out disparity averaging to explain high thresholds. The spatial distribution the disparity signals has a profound effect on sensitivity. Disparity detectors respond optimally to signals separated by at least 1 deg vertically and .5 deg horizontally.

This document is currently not available here.