Robust Image Corner Detection Using Local Line Detector and Phase Congruency Model

In this paper, a new robust corner detection method is proposed for detecting and localizing corners of planar curves. First, edges are extracted using a canny detector. Then, a local line detector is developed, and edge-pixels are labeled on the basis of the information provided by the local line detector. After egde labeling, the approximate location of the corners are determined. Finally, the minimum moments of the phase congruency information is used at a local window to determine the accurate location of corners. The advantage of the proposed method is that it does not involve calculation of the curvature and the threshold value. Experimental results demonstrate it is particularly effective for natural images, and possesses a high detection rate than the present methods.