Study On Infrared And Visible Image Fusion

With the rapid development of sensor technology, image fusion plays asignificant role in each area. Image fusion is to make plenty of images which arecaptured by different sensors from a scene into one. This image remains criticalinformation contained by the others. So, we got a complement and reliabledescription of the aimed scene.Infrared(IR) and visible image fusion is one of the important applications ofimage fusion technology. For an image can be divided into approximation globalinformation and fringe details in sub images at each scale by methods based onmulti-scale decomposition(MSD), which supplies the celerity and flexibility toevaluate, select and fuse the image information. This paper studies two typical imagefusion algorithms based on MSD, image fusion based on pyramid transform andwavelet transform. Meanwhile, we implement experiments fusing IR and visibleimages to validate these algorithms, then compare their fusing effects and improvethem.At first, this paper introduces the basic concepts and key steps of image fusionbriefly and describes image registration, which has a great influence on the effect ofpixel-level image fusion, in detail. Besides, we introduce two kinds of feature pointsdetection algorithm, SIFT and SURF, then propose experiments to exam theirreliability in IR and visible image fusion. The results turn out that these twoalgorithmn can find out the feature points in images effectively. However, there aredifferences between the IR and visible imaging principles and the features presentedin IR images and visible images are different from each other. The mismatch ratio ishigh. So it’s better to match the feature points manually.Next, in the researches of image fusion based on MSD, it proves the images,isof better quality obained by methods based on wavelet transform, especially bymethods based on daul-tree complex wavelet transform(DTCWT). What’s more, thecontrast of an image is proportional to the relative variation of the gray scale afterthe image is analyzed by wavelet. And with the scale increasing, at least the meanand variance of impulse noise and Gaussian noise linearly decrease. Because the IRimages have low definition and contain much noise which causes the degradation ofthe fused image quality. So we propose an enhanced image fusion method based onwavelet transform, which improves the contrast and IR information in the fused image effectively.At last, combining the image fusion theory in this paper with methods ofsoftware engineering, we designed and implemented an IR and visible image fusionsoftware. It functions well in the practical item.