Study On Homologous Image Registration And The System Design

Image registration algorithm as a topic in pattern recognition and computer vision is very important, in medical image analysis, computer vision, remote sensing image technology, an important field is widely used, which is one of the key technologies in image processing applications, due to the image element of diversity, registration efficiency of image registration, registration accuracy and adaptability have taller and more in-depth. Image registration algorithm based on registration technology can be divided into manual registration and automatic registration. In this paper, automatic registration is studied and introduced. Automatic image registration algorithm based on the image processing method can be divided into two kinds, based on the feature points of the image registration and region based image registration. This paper mainly discusses the advantages and disadvantages of these two algorithms and application areas, and made a corresponding classification introduced and improved. Image registration techniques, composition, can be understood as the contrast for the control points, or regional, by comparison, the pictures and to be image registration in the same object spatial location and spatial structure shows consistency, the figure in a meaningful role point or region is formed matching relationship, thus rendering the image registration after. The main work of this paper is:(1) the basic concepts, basic processes, common registration methods and basic principles of the feature-based registration method. For feature based registration method, including the feature points extraction, feature points matching and other basic processes, this paper also briefly introduces the extraction,matching principle and the basic operation. Feature extraction is a basic and key step.Therefore, the standard is put forward a Harris corner detection algorithm is used to extract feature points, and then the end of SIFT feature for rotation and scale of the deformation and for robustness to noise and the changes of the angle of view and extract the initial feature points, using random sample consensus algorithm(RANSAC) screening out those pseudo matching feature point pairs, to improve and enhance the image matching accuracy;(2) Secondly, in another aspect. This paper also on registration algorithm based on regional characteristics were briefly introduced and explained optimization, based on the regional characteristics of selected control points of a large quantity of reference advantages, but in order to avoid the complexity of the algorithm is high, the precision is bad, this paper presents a kind of small element surface pretreatment, and then extract the closed chain code matching algorithm of the second match of the series of operations. According to the experimental results, this method can avoid the disadvantages of traditional chain code method, and it can improve the stability and precision of the algorithm.