Performance Comparisons of Contour-Based Corner Detectors

摘要: Corner detectors have many applications in computer vision and image identification retrieval. Contour-based corner directly or indirectly estimate a significance measure (e.g., curvature) on the points of planar curve, select curvature extrema as corners. While an extensive number contour-based been proposed over last four decades, there is no comparative study recently detectors. This paper attempt to fill this gap. The general framework detection presented, two major issues—curve smoothing estimation, which impacts performance, are discussed. A promising compared using both automatic manual evaluation systems large datasets. It observed that while indirect estimation techniques more robust, direct faster.