Research On Infrared And Visible Image Fusion Method Based On Objective Extraction

Image fusion is a technology to integrate object or scene information from multipleimage signals into a single image. Through combing the complementary informationfrom multiple images, it enhances the availability, clarity and identifiability of thefusion image. This paper research the infrared and visible image fusion methods. Thevisible light images show is the directly observed result of the human eye to scenes, itsatisfy human visual characteristics, has high resolution and rich details, but it has apoor environment adaptability and the disadvantage of the weak anti-interference ability.While the infrared sensor reflects the thermal radiation of scene, it can generate imagesaccording to the temperature change and work well in bad environment such as dark,dusty and smoggy, etc. Because of the well complementary of infrared image andvisible light image, the fusion algorithms between them have been widely discussed.From the point which the objects information in infrared images are very obviouslywhile the visible images have more background details, this paper concentrates onfusion methods based on infrared image objective extraction.Firstly, this paper describes the basic principle of image fusion methods base onmulti-resolution analysis(MRA). Then introduce some MRA tools, including LaplacianPyramid Transform(LPT), Discrete Wavelet Transform(DWT),Stationary WaveletTransform(SWT) and Non-subsampled Contourlet Transform(NSCT). And theadvantages and disadvantages of them are analyzed.Through the analysis of image fusion algorithm from traditional Piella MRAframework, a novel fusion algorithm is put forward based on infrared image targetextraction in this paper. It combines infrared target extraction and MRA together. Firstly,the object extraction algorithm is adopted to segment infrared image into target imageand infrared background image. Then fuse the background of infrared image withvisible image by the method based on MRA to generate the comprehensive backgroundimage. Finally the infrared target image and the comprehensive background image isfused to get final fusion image.Two methods are put forward. Infrared and visible image fusion based on greyrelational analysis target extraction and NSCT and fusion method based on k-meansclustering and SWT. Respectively, two new infrared object extraction algorithm areproposed: In the first algorithm, the image grey theory is introduced to put forward a kind of infrared target extraction algorithm based on grey relational analysis(GRA);Inthe second algorithm, a new object extraction algorithm based k-means clusteringcombine with Canny edge detection is proposed. The simulation experiment show thatboth target extraction algorithm have good target extraction effectAfter the multi-resolution decomposition of infrared background image and visibleimage, a fusion criteria base on fuzzy logic is used to fuse the low frequencycoefficients. Cauchy membership function and Gaussian fuzzy logic fusion criterion areused to the two fusion methods, respectively. The simulation experiments show that thefusion rule based on fuzzy logic can improve the vision effect and objectiveperformance metrics of the final fusion image.Fusion algorithm based on infrared target extraction is proposed in this paper,through the simulation experiments show that the image target in the finally fusionimage is outstanding, the the finally fusion image has clear background good visualeffect. The preliminary research object has obtained. It also has certain referencesignificance to other image fusion algorithms.