The Dim Light And Infrared Images With Quasi-technical Studies

Image registration is an important topic in the field of image processing and is widely applied in remote sensing, computer vision, motion estimation, the medical image analysis and military night vision as well as other fields. It is the key step in fusion process.There are relative translation, rotation, zoom and even different ratio distortion between the images gathered from different sensors, different imaging modalities and different time.In order to actualize a comprehensive analysis and application, the image registration problems have to be solved first of all.This paper, the low-light and the infrared images, the principle and method of image registration are introduced. Three typical registration algorithms are focused on. Then propose and design a new registration method, which is combined of the image features and the gray mutual information, to improve the accuracy of registration and the application.In the paper, Moravec, Harris and Forstner feature extraction algorithm are analyzed. Three feature pint match algorithms are analyzed too. And through a large number of simulation experiments to determine the improved Forstner and the improved position similar to the triangle algorithm are the best algorithms to extract the feature point and to match them. Through studying the simplex search algorithm, simulated annealing algorithm, ant algorithm and considering to the registration we have done,â… choose the simplex search algorithm to be the final search algorithm. Finally use the MATLAB to achieve simulation, and evaluate the registration results on the value of mutual information.The innovation departments are:propose a new registration method which combines the gray mutual information and the image features, and improve the position similar to the triangle algorithm.