Border ownership from intracortical interactions in visual area v2

Area of Science:

• Neuroscience
• Computational Neuroscience
• Computer Vision

Background:

• Border ownership is critical for image segmentation and surface perception.
• V2 neurons signal border ownership, but this relies on information beyond their classical receptive fields.
• The precise mechanism by which V2 generates border ownership signals remains unclear.

Purpose of the Study:

• To investigate whether the V2 visual area can generate border ownership signals intrinsically.
• To model the intra-areal interactions within V2 that may underlie border ownership signaling.
• To determine if V2 requires top-down input or explicit labels to compute border ownership.

Main Methods:

• Development of a computational model of V2 neurons.
• Model neurons possess classical receptive fields, orientation tuning, and receive input from V1 regarding border location and orientation.
• Simulations focused on intra-areal interactions within V2 to generate ownership signals.

Main Results:

• The V2 model successfully generated border ownership signals.
• These signals arose from finite-range, intra-areal interactions within the model.
• The model reproduced physiological observations without needing explicit figure labels or top-down mechanisms.

Conclusions:

• V2 can generate border ownership signals autonomously through local computations.
• Intra-areal interactions within V2 are sufficient for computing border ownership.
• The model provides a basis for understanding V2's role in figure-ground segregation and suggests testable predictions.