A GRAY-LEVEL 2D FEATURE DETECTOR USING CIRCULAR STATISTICS

This paper presents a new method for corner and circular feature detec tion in gray-level images. It is based on the application of standard statistical techniques to the distribution of gradient orientations in a circular neighborhood of the prospective feature point. An evaluati on using standard procedures and a comparison with other approaches is presented. Results show the robustness of this method as compared to the other corner detectors analyzed. The main novelties are the possib ility of detecting points that are centers of circular symmetries, and discriminating between junctions, which are classified into corners ( two-edge junctions) and multiple edge junctions. (C) 1997 Elsevier Sci ence B.V.