Visual cortex inspired junction detection

Abstract

This paper proposes a visual cortex inspired framework to detect and discriminate junction points in a visual scene. Inspired by the biological research in primary visual cortex V1 and secondary visual cortex V2, banks of filters are invented to simulate the response of complex cells, horizontal lateral connections and angle detectors. Before junction detection, contour enhancement and texture suppression are performed in determining the sketch of a scene. The proposed algorithm can extract primal sketch of the scene and detect the response of junction points in the scene. Experiments prove the good performance on scenes with man-made structures. © 2011 Springer-Verlag.