Study On Fusion Algorithm For Infrared And Infrared Polarization/Vsible Images

Infrared imaging technology has many advantages and has been widely used in military,remote sensing,industrial detection,security and other fields.However,the traditional infrared imaging system gets the gray images of scenarios based on the objects' temperature difference and thermal radiation rate difference,which results in that these images have generally low contrast,blurry targets and poor overall vision.In some cases,there are some problems in infrared imaging system,such as detection failure,high miscarriage rate and tracking failure.In recent years,with the rapid development of various kinds of infrared camouflage technology,the problems above are particularly prominent.Visible image can reveal the background details of the observed scene well,and provide a good visual environment for infrared targets,which is convenient for target location,analysis,tracking,etc.Infrared polarization imaging technology can identify the target information effectively based on the different polarization information among different objects,and it breaks through the limitation that traditional infrared imaging technology depends on the objects' temperature difference,and can greatly improve the target recognition rate.Therefore,it has great importance of improving the performance of the imaging detection system to combine properly infrared image with infrared polarization image/visible image.Based on the study and in-depth analysis of the existing image fusion algorithms,several image fusion algorithms are researched for infrared and infrared polarization / visible images in this paper.The main contents and innovation points in my work include the following aspects:1.Combining the self-adaptive decomposition of bidimensional empirical mode decomposition and the flexible directional expansion of nonsubsampled directional filter banks,as well as the gray distribution features of infrared images,a fusion algorithm is proposed for infrared and visible images.2.Considering that infrared and infrared polarization/visible images contain common information and respective unique information,and sparse representation has the defect of losing high frequency information of image,an image fusion method based on bivariate bidimensional empirical mode decomposition and sparse representation is proposed.The proposed method own excellentperformance when processing infrared and infrared polarization/visible images.3.An improved multi-scale top-hat transform fusion algorithm is proposed by combining the multi-scale center around structural elements with top-hat transformation.Considering the gray distribution characteristics of infrared image,a method of obtaining the base image is designed by introducing the idea of Gaussian fuzzy logic,which can improve the fusion effect of infrared and visible images.4.Contraposing infrared image and infrared polarization images,there is a problem that these images are low contrast,fuzzy and gray concentrated,an improved multi-scale toggle contrast fusion algorithm is proposed to solve this problem.This algorithm can not only enhance markedly the contrast and clarity of fusion image,but also make the fusion image include rich details.