Infrared And Visible Image Fusion With Anisotropic Guided Filtering

Infrared and visible image fusion technology is one of the most widely used technologies in the field of multi-sensor image fusion.Because infrared and visible image sensor imaging mechanisms are different,infrared and visible images in the same scene have good complementary characteristics.Therefore,the target information in the infrared image and the background detail information in the visible light image are effectively fused.It can improve the recognition of the target,obtain a more comprehensive and accurate description of the scene,and be of great significance for subsequent applications such as target identification,detection and tracking.But at present,infrared and visible image fusion is prone to the loss of edge detail information,the existence of halo artifacts and false noise and other problems.Starting with the full acquisition of the important features of multi-source images and improving the fusion quality of infrared and visible images,this paper adopts a new multi-scale transformation tool based on edge-holding filtering technology-Anisotropic Guided Filtering.This paper discusses the fusion algorithm and edge retention filtering technology of infrared and visible image based on multi-scale transformation,and puts forward the corresponding solutions and algorithms.The main findings are as follows:The mainstream infrared and visible image fusion algorithms based on multi-scale transformation are discussed in depth,and the basic principles and advantages and disadvantages of these algorithms are analyzed.On this basis,in view of the problem that traditional multi-scale transformation can easily produce halo artifacts at the edge of the image,a multi-scale decomposition model based on anisotropic directional filtering is proposed.The model decomposes the source image into a basic layer containing large-scale changes and a series of detail layers containing small-scale detail information,which fully realizes the scale separation of image features.The final fusion results show that the decomposition model effectively retains the edge profile information and texture structure of the multi-source image.Based on the multi-scale anisotropy-oriented filter decomposition model,in order to further improve the contrast and clarity of fusion results and effectively improve the quality of fusion results,an initial weight graph construction algorithm based on phase consistency operators and contrast pixel significance is proposed.The algorithm uses phase consistency operators and Gauss filtering to calculate significant graphs,and then obtains the initial weight binary plot by comparing pixel significance.Experimental results show that the initial weight map obtained by the model can capture and extract more detail features and edge contour information of the source image,and at the same time suppress noise interference to a certain extent,which can help to deal with the initial weight graph more targetedly in the future.A new initial weight graph optimization algorithm based on anisotropy-oriented filtering is proposed for the initial weight graph obtained by most of the existing fusion algorithms during fusion,which may have noise or cannot be aligned with the image boundary.By optimizing the weight map with anisotropy-oriented filtering,the detail "halo" can be avoided while effectively overcoming noise interference.In the optimization process,the source image is used as the guide image,which maintains spatial continuity and further improves the fusion quality of the image.In order to verify the validity and applicability of the proposed algorithm,four multi-scale transformation fusion algorithms and four classical fusion algorithms were qualitatively and quantitatively analyzed on the Road Scene and TNO data sets,respectively.Experimental results show that the algorithm proposed in this paper is not only superior to other comparison algorithms in edge detail,background preservation and target integrity,but also effectively suppresses detail "halo" and artifacts,and has the characteristics of scale perception,which effectively improves the accuracy of fusion results.