A New Corner Detection Method Based on Differential Code

Detecting dominant points is an important subject in pattern recognition and computer vision. The paper sums up the traditional corner detection methods based on Freeman codes and presents a new way to detect corner. Firstly, the points are classified using the differential code and the exact corners and uncertain corners are got. Then the curvature of the uncertain corners is calculated, and the points whose curvature is greater than the threshold are candidate corner. At last the right corners based on the Moving Line method is got. Experimental results show that the proposed method is superior to the conventional methods. Besides, the proposed method can be carried out more quickly.