Abstract
Introduction: The general lay out of a scene influences shape perception, as is shown in shape contrast effects where the perception of one shape is distorted in the direction opposite to another shape. So-called cue combination models describe the perception of 3-D shape accurately. Such models are based on the assumption that the overall shape estimate is a weighted linear combination of the estimates derived from the individual cues and explain shape contrast on the cue level, before cue integration. However, curvature contrast effects for motion-, shading-and-texture and stereoscopically defined stimuli are all of the same order and magnitude. This suggests that these contrast effects are shape-based, not cue-based. In this study we investigated whether curvature contrast effects are cue-based or shape-based.
Methods: In one condition, both central and flanking paraboloids were defined in stereo. In another condition the flankers were defined by structure-from-motion, whereas the central test paraboloid was defined in stereo. We varied the curvature difference between the flanker paraboloids in the reference and test interval. Observers had to decide which of the two intervals contained the central paraboloid with the highest curvature. In this way we could determine the curvature contrast effect in all conditions.
Results: We found a consistent contrast effect in the stereo-stereo condition, in the motion-stereo condition it varied in strength for individual observers. In addition, we found that half of the observers that performed normally on the stereo-stereo condition barely perceived stereo depth when the flanker shapes were defined by structure-from-motion. This might have been due to conflicting binocular flatness cues in the structure-from-motion defined flankers.
Conclusion: Curvature contrast effects are at least partly shape-based, however, in some cases strong cue-conflict seems to prevent influence from other cues.