Contours drive distinct orientation selectivity in the human visual system

Area of Science:

• Visual perception
• Neuroscience
• Computational vision

Background:

• Orientation processing is crucial for visual perception.
• The specific image features driving natural scene orientation processing are not fully understood.

Purpose of the Study:

• To investigate whether human orientation judgments rely more on object contours or filter-based orientation energy.
• To determine how the human visual system encodes orientation in natural scenes.

Main Methods:

• Behavioral experiments measuring orientation judgments on image patches.
• fMRI data analysis using computational models (Photo-Steerable Pyramid, Line drawing-Steerable Pyramid, Contour-based approach) on the Natural Scenes Dataset.

Main Results:

• Participant judgments consistently aligned with contour-based orientation estimates, not filter-based ones.
• Models emphasizing contour structure provided a better explanation of neural fMRI responses.
• Distinct patterns of orientation preferences were identified in neural data.

Conclusions:

• The human visual system prioritizes contour information for encoding scene orientation.
• Object boundaries are more influential than filter-based orientation energy in natural vision.
• This challenges existing models of visual representation and emphasizes boundary importance.