Research Of Image Registration Algorithm Based On Image Feature

Today in the advancement of information, image fusion has penetrated into all fields in image analysis and processing which plays a decisive role. In order to obtain the same scene or object of a more comprehensive, accurate and reliable description of the source, image fusion uses pieces of registration information and comprehensively analyzes. Handling multi-source image registration problem is the fusion problem to be solved. Image registration is the preliminary work of the image fusion. The purpose of the image registration is to eliminate the differences between images. The level of registration accuracy directly determines the quality of image fusion.In a different time, using different sensors, on different perspectives and on different shooting condition, Image registration matches the geometric relationship of the two or more images. Its main purpose is to be removed or inhibited the geometric inconsistencies of reference images and moving image registration, including translation, rotation, scaling deformation. Feature-based image registration method has two main difficulties: the diversity of feature extraction and similarity computation complexity. In this paper, in view of these two issues, the main job is as the following: Firstly, algorithms of edge and corner extraction operators has been comprehensively analyzed and based on many tasks of corner extraction detection, operator of the improved Harris method has been proposed.Secondly, in image matching algorithm, the article analyzes a variety of similarity testing algorithms, comprehensively analysis of the a the advantages of alignment degree metric and use it in matching of corner feature point pairs , and based on the feature corner histogram obtain rotation angles, and calculate alignment degree of the feature points in image.Thirdly, the characteristics of the image information and the angle information has been combined with and new image registration methods based on corner features and the alignment degree of similarity of image has been proposed. In this way, only the extracted feature points have been calculated to reduce computation. At the same time, without the need for optimal searches corners has been matched in the process to avoid falling into local minimum.In this text, the corner detection, corner matching and eliminating of false matching points and affine transformation method have been analyzed and studied. Finally using MATLAB software, the computer show that the algorithm, and a comprehensive analysis of registration results and performance of the algorithm. Through the paper registration method it can be seen to achieve high registration accuracy, reliability and to be effective image registration.