Research On Image Fusion Method Of Infrared And Visible Images Based On ICA

Infrared and visible image fusion aims to capture thermal radiation and visible sensor to supplement texture detail information through infrared sensors,so as to obtain a fused image containing targets and more textures.The main goal of this thesis is to find an appropriate infrared and visible image fusion method,and obtain a fusion result that contains more useful information by fusing the two source images.Through the analysis of the current research status,it can be found that most of them are aimed at improving the saliency of the target,thereby ignoring the environmental information and the texture detail information of the target.Therefore,this thesis focuses on improving the texture details and environmental information of the target on the basis of guaranteeing the target,so that the resulting image after fusion is more in line with the human visual perception.First of all,this thesis improves the fusion strategy of infrared image and visible image design using Latent Low-Rank Representation(Lat LRR),and proposes a fusion method combining Lat LRR and Independent Component Analysis(ICA).First,use Lat LRR to represent the source image as low-rank,sparse and noise components;second,compare the entropy of the two source images to determine the image that contains more information,and use this as a reference;third,use the independent component analysis between the lowrank component of the reference image and the low-rank component of the non-reference image to obtain the main difference between the two source images;The components,sparse components and the reference image are fused to obtain the fused image.Experimental results show that compared with other fusion methods,such as CBF,Lat LRR,MLGCF,Fusion GAN,Res Net,and VGG,this method achieves better visual effects,and better evaluation index on MI and PSNR.Secondly,this thesis combines the theory of compressed sensing and Independent Component Analysis to improve a method suitable for infrared and visible image fusion.First,perform low-rank representation of the infrared image,and represent the infrared image as a low-rank image;secondly,use Compressed Sensing(CS)to sample and reconstruct the low-rank components of the infrared image;then,use the reconstructed image and the independent component analysis of the infrared image to obtain the main information part of the infrared image.Finally,the average weighting strategy is used to fuse the main information in the infrared image and the visible image to obtain the final result image.This method has been tested on the UNcamp and Solider images datasets.The experiments show that the results obtained by this method have clear texture.The results obtained by this method have clear texture,better spatial frequency SF evaluation index on the Uncamp dataset and VIF on the Solider-Shooting dataset.Based on the basic theory of image fusion,this thesis analyzes the research process and proposes an infrared and visible image fusion method combining Lat LRR and ICA,and combines the theory of compressed sensing to improve the infrared and visible image fusion strategy.Finally,the shortcomings of the current method and the direction of future improvement are discussed.