Multi-sensor Image Fusion Technology

Multi-sensor image fusion is a technique with one fusion arithmetic to combine the information of multiple images in the same scene from different sensors. Because this technique could use the information of different channels effectively, the researchers paid more attention on it. Recent years, image fusion technique has been improved a lot. At the same time, it has been used in many field successfully.To meet the needs of military automatic object detection, this paper focus on multi-sensor image fusion to increase object detection probability, object recognition ability and object orientation precision. The research includes the following parts.Firstly, the conception, background and development of image registration was discussed in this paper, as well as the classes and basic principle of image registration arithmetic. On the base of these analyses, the paper makes some amelioration and innovation. Using the characteristics of infrared image and visible image, a arithmetic of image registration which is based on speeding edge distill and maximizing uniform method is put forward. The new arithmetic ignores the spectrum of images, and just pays attention on edge traits of images, the method could be used in the registration of different wave band images, and stops the interference of image brightness to the registration of two or more images. At the same time, this new arithmetic has could registed the images efficiently.Secondly, the existing image fusion methods and their capabilities was studied in this paper, and then gets an arithmetic that based on regional features measurement and Laplacian Pyramidal decomposition. Because of using the inside details and structure information instead of single pixel brightness information, the arithmetic has a better image fusion effect. As a result of the characteristic of Laplacian Pyramidal, the operation efficiency of the new arithmetic is high. In addition, the influence of the fusion results with different arithmetics was compared in this paper, as well as the differences of different arithmetic, and then proves that the new arithmetic is better than the others.Thirdly, a rationed appraisement to the image fusion effect in some methods was got in this paper. The author suggests that the entropy is the best parameter to appraise image fusion. We could use this parameter to appraise images which are fused in different methods. The results of experiments prove that the appraised method and criterion are reasonable and effective. Forthly, in the mock military scene of object detection, the arithmetic of image registration which is based on speeding edge distill and maximizing uniform method to realize multi-wave band registration was used, and then gets the image fusion based on regional features measurement and Laplacian Pyramidal decomposition method. The results of the experiments are very good.In conclusion, this paper focuses on the principle, measure and estimate of multi-sensor images registration and fusion, and gives a full-scale and systemic research. By a great deal of experiments, some important conclusion was gained in this paper, which could provide useful suggestion to the research of image fusion. At the end, the results was summarized, and this paper discussed the contribution of this research and provided some suggestions for future research. Limitations of this research were also discussed.