A neural network model of contours extraction based on orientation selectivity in the primary visual cortex: applications on real images

The capacity of the primary visual cortex (Vl) to extract salient contours from real black-and-white images is studied using a neural network model of information processing in V1. The model includes the input from the lateral geniculate nucleus, arranged according to the preferred orientation through a Gabor function, a feedforward inhibition from inhibitory interneurons and lateral connections (both excitatory and inhibitory) from the other cortical cells (feedback mechanism). Intracortical excitation is arranged according to experimental data, in order to implement the Gestalt proximity and good continuation criteria. Intracortical inhibition realizes a competitive mechanism among neural groups, to eliminate noise. Simulation results, performed on black-and-white images, demonstrate that the network can easily extract salient contours, by suppressing zones of constant luminance and isolated noise, with an acceptable settling time (30-40 ms). The role of intracortical synapses was also analyzed: too excessive extension of intracortical inhibition can suppress small contours, on the other side too reduced intracortical inhibition can cause the appearance of superimposed noise in the image.