Research On Infrared And Visible Image Fusion Algorithm Based On Multi-Scale Decomposition

Image fusion technology refers to the fusion of information from two or more source images onto one image,so that the fused image contains more comprehensive and useful information from the source image.In the field of image fusion,the fusion of infrared and visible light images is a key focus of current research.The purpose of fusion is to effectively integrate the significant important target information in infrared images with the rich texture details in visible light images,in order to obtain a clear image with rich information that meets practical needs.However,currently existing image fusion methods have problems such as unobtrusive targets,high artifacts,low contrast,poor visual effects,and loss of texture details in fused images.Based on theoretical research on image fusion methods,this article proposes three infrared and visible light image fusion methods based on multi-scale decomposition to address the above issues.The main research content is as follows:(1)Aiming at the problem that the salient targets in the fused image are not prominent and contain many artifacts,a hybrid multi-scale decomposition model based on sub window variance filter and Gaussian filter is proposed for infrared and visible image fusion.Firstly,the source image is decomposed into basic layers,fine structure layers,and coarse structure layers using this decomposition model.Secondly,the basic layer adopts a weighted average fusion rule based on visual saliency mapping;The coarse structure layer combines the advantages of convolutional neural networks and adopts fusion rules based on Laplace pyramid and local similarity;The fine structure layer adopts a summation strategy for fusion.Finally,the fused base layer,fine structure layer,and coarse structure layer are added and summed to obtain the fused image.The experimental results show that the fusion algorithm can better preserve significant target information in the source image,and the fused image contains fewer artifacts.(2)Aiming at the problems of low contrast and poor visual effect in the fused image,a hybrid multi-scale decomposition model based on L1L0 and rolling guide filter combined with Gaussian filter is proposed for infrared and visible image fusion.Firstly,obtain contrast enhanced images using the HLIPSCS method.Secondly,the contrast enhanced image is decomposed by L1L0 to get the base layer and detail layer,and the base layer is decomposed by rolling guide filter and Gaussian filter to get the detail layer and final base layer.Then,the final base layer adopts a weighted average fusion rule based on visual saliency mapping;The detail layer obtained through L1L0 decomposition adopts fusion rules based on Laplace pyramid and local similarity;The detail layer obtained by rolling guidance filter combined with Gaussian filter decomposition adopts the fusion rule based on weighted least square optimization.Finally,the fusion results of the final base layer and two detail layers are added and summed to reconstruct the fused image.The experimental results show that the fusion algorithm achieves better contrast and visual quality of the fused image,making it suitable for human observation.(3)A fusion method for infrared and visible light images based on visual saliency mapping and local gradient energy is proposed in a multi-level latent low rank representation framework to address the issue of texture detail information loss in fused images.Firstly,the source image is decomposed into a low rank part and multiple detail parts through multi-level latent low rank representation decomposition.Secondly,the low rank part is fused using a weighted average fusion rule based on visual saliency mapping,while the detail part is fused using a column by column fusion rule based on local gradient energy.Finally,the fused image is reconstructed through multi-level latent low rank representation inverse transformation.The experimental results show that the fusion algorithm can preserve richer texture details and further improve the quality of image fusion.