Corner Detection Based on Morphological Disk Element

Corner detection in image processing plays an important role in matching, representation, recognition, registration, and fusion. As one of the template based approaches, the morphological detector has been an interesting research field. But many problems are still under research, such as the element selection, high computation cost, and inaccurate representation of detected corners. In this paper, we propose a new morphological detector, which uses simple symmetric disk element in corner detection to avoid element rotation and improve the running efficiency. As a template based approach, this detector can accomplish the detection without any manual operation. Besides, different from other template based detectors, the detected corner in this approach can be represented with a pixel point, which is more accurate than area representation. The important character of this detector is the ability in estimating corner angle and orientation during the detection. These attributes of the corner have a wide applications in image processing. The performance of the new detector is tested on a set of standard images. Six other detectors are involved to provide comparative performance information. The experiment results prove the superiority of the new detector.