Image reconstruction explains the properties of horizontal connections in V1

Image reconstruction using multi-resolution scheme is proposed as a new computational role of horizontal connection (HC) in V1. Based on the new role of HC in V1, a neural network model of V1 is naturally derived by applying steepest descent method to image reconstruction. The properties of HC of the model are consistent with neurophysiological experimental results. Numerical simulations show that the model is effective to detect edges under noisy condition and it has an ability of contour integration.