Temporal evolution of pattern disparity processing in humans.

Area of Science:

• Neuroscience
• Computational Vision
• Visual Perception

Background:

• Stereo matching is fundamental to depth perception.
• The visual system's mechanism for computing 2D pattern disparity remains incompletely understood.
• Previous models suggest disparity is extracted within orientation-selective channels.

Purpose of the Study:

• To investigate the neural computation of 2D pattern disparity in human subjects.
• To determine how the visual system combines 1D disparities to form 2D pattern disparity.
• To explore the role of orientation-selective channels in stereo matching.

Main Methods:

• Measurement of reflexive vergence eye movements in human subjects.
• Presentation of stereo plaids and gratings with varying spatial frequencies and contrasts.
• Analysis of neural computation timing (10-15 ms) for pattern disparity.

Main Results:

• Compelling evidence for combining 1D disparities within orientation-selective channels to compute 2D pattern disparity.
• Pattern disparity computation persists under conditions of perceived transparency.
• Failure to compute pattern disparity when grating spatial frequency and contrast differ significantly, suggesting limitations beyond simple transparency.

Conclusions:

• A narrow-band implementation of the Intersection of Constraints (IOC) rule, preceded by cross-orientation suppression, underlies pattern disparity extraction.
• The visual system effectively combines orientation-based disparity information for stereo vision.
• Stimulus properties like spatial frequency and contrast critically influence the computation of pattern disparity.