Häkkinen J, Liinasuo M, Kojo I, Nyman G, 1996, "Is stereoscopic capture possible with three-dimensionally slanted or curved illusory surfaces?" Perception 25 ECVP Abstract Supplement
Is stereoscopic capture possible with three-dimensionally slanted or curved illusory surfaces?
J Häkkinen, M Liinasuo, I Kojo, G Nyman
The depth of disparate illusory contours affects the perception of background patterns that are enclosed by these contours. If the pattern is repetitive, the depth of the illusory surface is attributed to texture elements even though the texture is in zero disparity. Previous results have suggested that this phenomenon, called stereoscopic capture, is possible only with frontoparallel surfaces (Ramachandran, 1986 Perception & Psychophysics 39 361 -- 373). We hypothesised that the disruption of stereo capture with three-dimensionally slanted or curved surfaces has been due to the impossibility of consistent rematching of background texture elements. If the texture is designed in such a way that adjacent elements can be rematched to form a surface that is consistent with the three-dimensional structure of the illusory surface, the capture should be possible with complex surfaces.
We investigated our hypothesis by showing three-dimensionally slanted and curved surfaces to subjects and changing the pattern of the background texture. According to our results a texture of a constant period can be captured only by a frontoparallel surface; however, if the texture is designed to be consistent with the three-dimensional structure of the complex surface, the texture elements that initially form a frontoparallel surface form a complex surface when they are captured.
We conclude that (a) stereoscopic capture is possible with complex three-dimensional surfaces; (b) the possible discrete matches within the background texture determine the possible three-dimensional forms that can be captured and therefore stereoscopic capture should not be characterised only as depth interpolation (Mitchison and McKee, 1987 Vision Research 27 285 -- 294) between large image elements.