Selective biasing of stereo correspondence in an ambiguous stereogram

Area of Science:

• Visual neuroscience
• Computational vision
• Perception psychology

Background:

• Stereoscopic perception research has not fully elucidated feature matching between the eyes.
• Previous studies often used unambiguous stimuli, limiting the complexity of the correspondence problem.

Purpose of the Study:

• To investigate the stereo correspondence problem with a novel, ambiguous stimulus.
• To contrast the nearest neighbor matching rule with the nearest disparity rule.
• To quantify the visual system's preference between these matching constraints.

Main Methods:

• Developed a novel 'wallpaper' inspired stimulus to create perceptual ambiguity.
• Manipulated contrast of adjacent image features to create conflicting matching cues.
• Quantified the visual system's preference for specific matching rules.

Main Results:

• Demonstrated a quantifiable preference for nearest disparity matching over nearest neighbor matching.
• Showed that the magnitude of this preference depends on the scene's disparity levels.
• Results align with the hypothesis that the visual system minimizes local disparity variations.

Conclusions:

• The visual system prioritizes minimizing local disparity differences, favoring the nearest disparity rule.
• This finding has implications for computational models of stereo vision and the role of prior constraints.
• The study provides new insights into solving the complex correspondence problem in stereopsis.