Stereo vision and isoluminance.

Abstract

Recent experiments have shown that stereo depth is given by fusion of illusory ('cognitive') contours. They occur across quite large gaps in figures, when these gaps are unlikely and form the shape of a probable (nearer) masking object or masking feature. Implications are that: (a) clearly defined contours and regions of brightness difference can be produced as postulates from sensory evidence, which may be surprising absence of stimulation; (b) each eye-system can derive its own postulates, or hypotheses, which (c) can be combined to give stereo vision. It has been shown that random-dot stereo depth does not occur when there is colour contrast but no brightness difference between the dots and their background. This we have confirmed by using a new technique for producing isoluminant pictures, of any complexity, with exact registration for any two colours. With this technique, we find large displacements of narrow borders bounding regions that are shifted across isoluminance. These displacements, which are clearly seen as movements, occur with or without colour differences. The direction of shift depends on whether the narrow border is light or dark. It is found that these dramatic shifts do not - when produced in opposite directions to the two eyes and fused - produce stereo depth. It is concluded, following Julesz's paradigm, that these contour displacements have their neural orgin not retinally, but after stereo fusion. Experiments combining the 'cognitive contours' stereo depth with isoluminance are described.