Research On Infrared And Visual Image Fusion Method

With the development of modern imaging technology and the usage of the corporation of multiple imaging methods, image fusion methods have been developed to process the image data more efficiently, which were applied to military, remote sensing, medical imaging, public security, industrial, transportation and so on. Image fusion is important branch of data fusion. It integrates the information of multiple images to get a more effective single image and maintain the key features of the source images. Up to now, there were great many kinds of methods proposed for image fusion, each has its own feature and practicability. In this paper, we researched infrared and visual image fusion methods based on corporation of multiple wave band imaging.In this paper, different image fusion methods and the fusion related problem are discussed. The analysis and comparison of the existed multi-scale image fusion methods were presented based on multi-scale fusion strategy. Multi-scale image fusion method based on region saliency analysis and method based on dynamic saliency analysis are proposed according to the research on saliency detection methods and the characters of infrared and visual image fusion. After that, we extended our research to other applications related to multiple wave band imaging.We get region saliency map for infrared image firstly for region saliency based fusion method to distinguish the target region, background region and medium region. Fusion strategy is formulated integrating saliency value and local area analysis. Then we fuse images on multi-scale domain based on the strategy. Experiments show that our method has some advantages on both keeping infrared features of target region and details of other regions better than traditional methods.The difference of inter-frame of sequential images represents dynamic information on temporal domain. We get the dynamic saliency distribution by integrating spatial and dynamic information of the original infrared frames. Then we fuse sequential image frames based on dynamic saliency. Experiments show that the proposed dynamic fusion methods can both keep spatially salient information and dynamic information and then is better than the static fusion methods.Besides, we did some preliminary research on infrared image enhancement and multi-scale image fusion strategy based on saliency detection. It shows that the research has some effect in the experiments.The points of innovation of this paper are as follows. For the first, we proposed an region saliency based image fusion method which formulate fusion rule based on region saliency and thus has a better result than traditional methods; For the second, we proposed an dynamic saliency based sequential image fusion method which combines spatial information and dynamic information of intra frames and can be effectively used for fusion of sequential images; For the third, we proposed an infrared enhancement method based on edge preserving filter and saliency detection, which modulates image gray levels according to saliency value and can enhance the contrast of targets.