A feedback model of figure-ground assignment

Area of Science:

• Computational Neuroscience
• Visual Perception
• Cognitive Science

Background:

• Understanding figure-background segregation is crucial for visual perception.
• Existing models often struggle to integrate bottom-up and top-down processing.
• The interaction between the ventral and dorsal visual streams is key to visual processing.

Purpose of the Study:

• To propose a computational model explaining the unified perception of figure and background.
• To elucidate how bottom-up and top-down signals are combined in visual perception.
• To investigate the role of ventral and dorsal stream interactions in figure-ground segregation.

Main Methods:

• Development of a computational model based on ventral and dorsal stream interactions.
• Utilizing a recurrent surface network to form surface representations.
• Computer simulations to test the model's predictions against established principles.

Main Results:

• The model correctly assigns figural status based on properties like size, contrast, convexity, and orientation.
• It simulates principles of lower region processing and top-bottom polarity.
• The model explains how exogenous and endogenous attention can reverse figural assignment.

Conclusions:

• The proposed model offers a unified framework for figure-background perception.
• It highlights the importance of dorsal stream saliency computation and surface network dynamics.
• The model demonstrates object-based attentional selection through neural activity spread.