Cognitive Vision Inspired Corner and Vertex Detection Based on Endstopped Neurons

This paper presents a visual cortex inspired method for corner and vertex detection. Endstopped cells in cortical area VI, which combine outputs of complex cells tuned to different orientations, serve to detect line and edge crossings (junctions) and points with a large curvature. This property is used for corner and vertex detection in this paper. We utilize the endstopped neurons model of Dobbins, and apply a novel scheme to select the "cornerness" measurement, so the edges and textures can be rejected automatically. Experimental results with synthetic images and natural images show that the proposed method detects corners and vertices more efficiently and accurately, performs better in the presence of noise than Harris and Susan detectors, and the detections keep stable under image distortions.