A CORF computational model of a simple cell with application to contour detection

Abstract

We present a computational model of a simple cell with application to contour detection. The response of a combination of receptive fields (CORF) model is computed by combining the activity of model lateral geniculate nucleus (LGN) cell using Difference of Gaussians (DoG) function with an AND-type function. Main conclusion: 1. The proposed CORF model is a computational model of a simple cell.2. Anatomically speaking, CORF is a more realistic model than the Gabor function model.3. Besides orientation selectivity, CORF exhibits contrast-invariant orientation tuning and cross-orientation suppression. The Gabor function model fails to produces these properties.4. CORF outperforms the Gabor function model in contour detection tasks with high statistical significance (p less than 10 -4 ). Contour detection is presumably the primary biological role of simple cells.