Illusory contour processing in early visual areas: a modeling approach
In this thesis, a computational model for illusory contour processing in early visual areas is proposed. In the model, image features are first measured in model V1. Then the resulting activities in V1 are fed into V2 cells with non-classical bipole property, which was modeled as the result of activity-dependent synaptic modification. The integrated illusory contour information in model V2 is then delivered to model V1 through excitatory feedback connections. A mechanism of unbalanced feedback synaptic weights to orthogonal domains was proposed to explain both V1 orientation reversal pattern in illusory context and excitatory feedback effects from V2 to V1 in real context. This context-dependent activation pattern in V1 and V2 might be the basis of illusory contour perception. Simulation of the model was provided and the result was consistent with existing physiological and psychophysical data.