Research On Multi-source Image Fusion

Multi-source images according to different sensors can be divided into optical image, infraredimage and radar image. As the different imaging mechanism, limited working time and thesurrounding environmental factors, single image often can not fully describe the scenes information.Fusing the optical, infrared and radar images into an image can complement each other and improvethe accuracy in the military.A large number of image fusion research is studied based on the data source of optical, infraredand radar images using the traditional image fusion technologies and NSCT transform fusion rules.Research work can be divided into two parts:First, optical, infrared and radar images parallel fusion algorithm based on the NSCT transform isstudied. Parallel fusion algorithm uses the same approach to process optical, infrared and radar images.Parallel algorithm has uniform transform and simple features.Second, optical, infrared and radar images serial fusion algorithm based on different integrationstrategies is studied. Serial fusion algorithm fuses optical and infrared image firstly, and then thefusion image is fused with radar image. Serial fusion algorithm has strong specific and targetedfeatures.The main findings of this research are stated as follows:1) Optical, infrared and radar imagesparallel fusion algorithm based on NSCT transform fusion image can not brightly display imagedetails.2) Optical, infrared and radar images serial fusion algorithm based on different integrationstrategies, first, optical and infrared images are fused using pixels maximum rule; and then the fusionimage is fused radar image using NSCT flexible transformation rules which have the features ofmaximum low-frequency coefficients and largest regional energy high-frequency coefficients. Fusionimage reflects the scene clearly and brightly with high optical and infrared identification and does notignore radar image. Serial fusion algorithm is the better solution of optical, infrared and radar imagefusion which can get a high level of comprehensive image.