An Improved Image Corner Matching Approach

Accurate and reliable image corner matching plays an important role in various applications in computer vision, such as motion tracking, object recognition and 3D reconstruction. This paper proposes an improved image corner matching method based on both blob information and local neighbor similarity. The similarity measurement between two interest corner points is based on corresponding SIFT information, and can therefore help to achieve robust corner matching against rotation and scale changes. Experimental results demonstrate that the improved method is effective for image corner matching with significant rotation and scale changes as well as other common image corner matching conditions.