Research On Edge-Based Image Registration Method

Image registration is a fundamental problem in image processing, which is a process to match two or more images of the same scene taken at different times, from different viewpoints, or by different sensors. Image registration is an important technique for a great variety of applications such as medical image analysis, computer vision, remote image data analysis, image information fusion, and the recovery of the 3–D characteristics of a scene.The edge-based image registration method is common used technique in the image registration. Based on the review of the current image registration methods, the disadvantages of those methods in the applicability and rapidity are disclosed. In this paper We aim to enhance the applicability of image registration method and speed up image registration process.There are two tasks which need to be handled during an image registration process. That is, the feature extraction and correspondence establishment. In this thesis several problems are discussed: Firstly, by the analysis of wavelet multi-scale edge detection principle, and different regulation of wavelet transformation to the noise and edge, we bring up wavelet multi-scale product the edge feature point extract algorithm which can availably remain the orientation of feature point. We can easily estimate the orientation difference between two overlapping images in next step. Secondly, we discover and analysis the characteristic of some feature matching measure such as cross–correlation and align metric. These feature matching measures can tolerate rotating variation at some degree. Thirdly, a so-called "feature sub-image" is presented. We can select some feature sub-image from the image by gray histogram. The method can significantly reduce the time of feature point extraction and feature matching and speed up the registration process. Finally, we make use of the above method, and present the edge-based rapid image registration approach. Compared with previous image registration method, the proposed scheme is a great improvement in rapidity as well as applicability for the image registration problem.