Automatic Image Registration Based on Natural Characteristic Points and Global Homography

Image registration based on the features of an image is a major research direction. It is common sense that there are many problems in traditional feature registration including a large amount of computations, high complexity and low accuracy. So, an auto image registration method based on natural characteristic points and global homography is proposed in this paper. In this method, firstly, the feature points are detected using Harris feature detector. Then the one-to-one matching points are obtained through rough matching and refine matching. Finally, image registration can be achieved through the homography matrix calculated by matching corner points. Experimental results demonstrate its fewer computations, high accuracy, insensitive to noise and illumination with good robustness. This study presents a new idea on image registration with a mutual rigid transformation relation.