Bayesian models of binocular 3-D motion perception.

Area of Science:

• Visual neuroscience
• Computational modeling
• Perception psychology

Background:

• Three-dimensional (3-D) motion perception studies reveal systematic biases in perceived trajectory angles.
• Existing models focus on motion-first or stereo-first processing.

Purpose of the Study:

• Investigate predictions from Bayesian models for 3-D motion perception.
• Determine the source of perceptual bias in 3-D trajectory perception.

Main Methods:

• Developed Bayesian models incorporating stochastic representations of monocular velocity and binocular disparity.
• Utilized a binocular viewing geometry with assumed noise in inputs and a prior for 3-D motion.
• Conducted two psychophysical experiments to test model predictions.

Main Results:

• Bayesian models generated testable predictions for perceived trajectory angle and velocity.
• Experimental results indicated that binocular disparity processing, not motion processing, is the source of perceptual bias.
• Systematic biases in perceived 3-D trajectory angles were observed.

Conclusions:

• Binocular disparity is the primary contributor to perceptual bias in 3-D motion perception.
• Bayesian frameworks provide a robust method for modeling and understanding 3-D motion perception.
• Further research can refine these models to better explain visual perception mechanisms.