Finding a series of efficiency interest points was important to feature point matching. In the paper we review the approach for detecting interest points. These approaches include the Harris detector and Susan detector. Then propose a novel approach for detecting interest point on boundary base on curvature. We extract the boundary with the edge detect base on Zernike moment and computer the curvature of every edge pixel gray. The candidates of interest points were detected based on curvature threshold. The experimental results show good capability of detection and localization.
展开▼
