The Study Of Infrared And Visible Light Image Fusion Algorithm

Infrared and visible image fusion is the technology that integrates information of multiple images by infrared sensors and CCD sensors of the same scene to generate more complete and accurate description of the scene. The fuse image with more object information is easier to identify.This dissertation mainly aims at the research of infrared and visible light image fusion algorithms based on the multiscale decomposition.In order to solve the issues of existed fusion algorithms which do not take the intrinsic characteristic of the source images into account,the priori information such as the imaging mechanism and the imaging characteristic of infrared image and visible light image has been deeply analyzed in this dissertation. The main thesis work and innovations as follows:1. For wavelet transform has many defects and ordinary contourlet transform is not shift-invariant, we select the nonsubsampled contourlet transform as source image frequency domain transform. Comparing to wavelet transform and ordinary contourlet transform, the nonsubsampled contourlet transform is a true two–dimensional transform that can capture the intrinsic geometrical structure and has been applied to many tasks in image processing for it has a high degree of directionality and anisotropy.2. To solve the fuzzy phenomenon of fusion images, the original images are decomposed by the nonsubsampled contourlet transform, the impact of low-frequency information to the image clarity is analyzed, and an adaptive image fusion algorithm of infrared and visible images is proposed. Both the change and uniformity of the image low-frequency are considered, as the basic of ensuring the clarity without the edge distortion, the defects of low contrast, high distortion is resolved effectively. Experiment results show that compared with the traditional method based on the Nonsubsampled contourlet, the entropy increased by 6.8%. 3. The traditional image fusion of Infrared and visible light images neglects the background differences and the targets complementary, resulting the poor clarity or week identifiability. According to the characters of infrared and visible light images, an algorithm based on prospect recognition is proposed. The infrared target region and visible corresponding area are fused by NSCT after the infrared target region is segmented by maximum between-cluster variance, the final fusion image is obtained by combing the target image and the background information of visible light image. The fused image exhibits good infrared target features as well as clear visible background, therefore, the visible effect and information content are improved maximally. Experiment results show that compared with the traditional method based on the Nonsubsampled contourlet, the entropy increased by 9.7%.The fused image can provide useful information for further computer processing, for example, image segmentation, object recognition, object detection, battle damage evaluation and understanding, and so on.