From stereogram to surface: how the brain sees the world in depth

Area of Science:

• Neuroscience
• Computational Vision
• Visual Perception

Background:

• Random-Dot Stereograms (RDS) reveal how binocular disparity generates conscious 3D surface percepts.
• Existing models explain disparity detection but not the generation of conscious 3D surface percepts from RDS.
• Stereograms demonstrate surface completion for occluded regions and figure-ground separation.

Purpose of the Study:

• Propose a neural network model for generating 3D surface percepts from RDS.
• Explain how visual cortex circuits interact to create conscious 3D surface perceptions.
• Unify explanations for stereopsis, 3D figure-ground separation, and surface completion.

Main Methods:

• Developed a neural network model based on interactions between visual boundary and surface representations.
• Investigated interactions between layers 4, 3B, and 2/3A in V1 and V2 for stereopsis.
• Modeled interactions between V2 perceptual grouping laws and V1/V2/V4 surface filling-in operations.

Main Results:

• The model predicts how layered visual cortex circuits generate 3D surface percepts.
• Computer simulations demonstrate the model's explanation of RDS-induced 3D surface percepts.
• The model successfully unifies data on stereopsis and 3D figure-ground separation/completion.

Conclusions:

• The proposed neural network model provides a unified explanation for 3D surface perception.
• Interactions between boundary and surface representations, following complementary rules, are key to consistent 3D visual percepts.
• The model clarifies the roles of specific visual cortex layers and operations in creating conscious 3D surface awareness.