Vergence eye movement control and multivalent perception of autostereograms

Area of Science:

• Computational neuroscience
• Computer vision
• Ophthalmology

Background:

• Stereo-vision relies on the correspondence problem for depth perception.
• Autostereograms present 3D information within 2D patterns, challenging visual systems.
• Automatic vergence eye movements are crucial for maintaining binocular fixation and depth perception.

Purpose of the Study:

• To introduce a dynamical model for automatic vergence eye movement control.
• To present a complete model for stereo-vision by integrating binocular neurons and vergence control.
• To apply the model to autostereograms and evaluate its performance against human perception.

Main Methods:

• Development of a dynamical system of binocular model neurons to solve the correspondence problem.
• Implementation of an automatic vergence eye movement control algorithm that adjusts image segments of interest.
• Testing the model's efficacy on computer-generated autostereograms.

Main Results:

• The model successfully adjusts image segments based on momentary observer interest with a minimal disparity search range.
• Application to autostereograms yielded results in full agreement with human multivalent 3D perception.
• The model exhibits a rapid phase transition to 3D perception, similar to natural vision.
• The algorithm demonstrated robustness against noise and eliminated the need for sparse depth map interpolation.

Conclusions:

• The proposed dynamical model provides a comprehensive solution for stereo-vision and automatic vergence control.
• The model accurately replicates human perception of 3D information from autostereograms.
• The algorithm's robustness and efficiency make it a promising tool for advanced visual processing applications.