The Research About Registration In Image Fusion Technology

With the constant development of computer vision technology, image fusion of image registration techniques have been widely used in medical, satellite images, remote sensing, military, and many other areas of application. Image registration is for a more comprehensive understanding of the same thing's information, so matching the same thing's multiple images. These images can come from different time, different equipment or different camera angles. Image registration not only is a key step in image fusion, but also is the basis for further processing image. Though the algorithms of image registration technology have a lot, but there is no single method is common to apply to all of the image registration. This study is to use the improved algorithm which is applied at different angles on the image registration method to enhance the speed and accuracy of the image registration of different angles and lay the foundation for further image fusion. The different angles image registration process is due with the images that are taken from the same target object at different angles.The main work of this paper is as follows:The first is improved method in detection the feature which is based on the Harris feature detection method. It adds to scale and affine invariant features to test image features. Experiments show that the improved method can not only find stable features Point, and to find feature regions.Second, the SIFT method is added to the treatment of noise to reduce the noise impact which in order to better describe the feature regions. Again based on RANSAC algorithm is improved to find errors match point faster and more accurately and delete it. Experimental results show that the improved algorithm is feasible and improve the speed and accuracy of the algorithm.Finally, the feasibility of the improved algorithm does experiment, image registration methods different angle experiment by using MATLAB software. The experimental results are analyzed and the results showed that the improved algorithm is feasible and can improve the image of different angles Registration accuracy and speed.