Research On The Region-based Fusion Algorithm For Infrared And Visible Images

The fusion of infrared(IR)and visible light(ViS)images is a research focus in machine vision,and has been widely applied in many fields such as remote sensing,night vision,video surveillance,military defense,and so on.The imaging resolution of visible sensor is superior and the detail of the edges and texture of objects are clear.Also,it has strong anti-interference capability.It's a pity that it is easily influenced by natural conditions.Imaging system of infrared sensor can penetrate through surface features such as smog and forest.Its subject goal is comparatively clear.It is more important that it can work perfectly at day and night.So the fusion of visible and infrared images can improve reliability of objects detection,reconnaissance,recognition,tracking,and achieve a more accurate,complete and reliable description of the same scene.At present,the image fusion technology is still not mature enough.There are some problems such as poor in details and robustness,and bad visual effect.This thesis introduces the basic theory of image fusion method,mainly studies the fusion method based on the target area.It proposed two novel region-based fusion algorithms for IR and ViS images.Experimental results show that the proposed method also can be further used in tasks dependent saliency detection and subsequent applications such as content-based image retrieval.The paper has carried lots of studies from two aspects,the main research work and contribution are as follows:1.We present a novel fusion algorithm for infrared and visible images based on Non-Subsampled Contourlet Transform(NSCT)and target area.Compared with other fusion methods,the Non-Subsampled Contourlet Transform(NSCT)is effective and efficient.It has the multi-resolution analysis,time-frequency local analysis,directional and anisotropic properties.The paper proposed a method of region based on NSCT transform,combining feature level and pixel level and increasing information of fusion image.Well solves the related problems of poor visual effect and the fused image details not enough through NSCT.By conducting experiment,we compare our method to others using several metrics and results show that the new method can improve the quality of the fused image and retain a better visual effect and can enhance the contrast,sharpness and detail information of the fused image.2.A novel region-based fusion algorithm based on interest points and saliency map is presented.Using the visual model of the traditional method for target extraction,it can highlight target area at the same time,also introduced some background region.Combining convex hull of interest points and saliency map,we can more accurately extract the target area.At the same time,it has better versatility and stability than the image segmentation algorithm.Different fusion rules for the object region and the background are employed respectively.This strategy makes the saliency result more robust and accurate to the object.Finally,Experiment results show that the proposed algorithm has good performance.