Research On Infrared And Visible Image Fusion Base On Sub-window Variance Filter

The image generated by a single sensor usually has only one aspect of information,and cannot achieve a complete and accurate description of the scene information,so the image fusion technology came into being.The infrared image mainly relies on the thermal radiation of the object itself to image,highlighting the thermal target hidden in the background,which is not affected by lighting conditions and weather,but has low contrast and rich texture details.Visible light images are imaged by reflecting visible light.Texture details and contrast are more suitable for human visual perception,but the imaging effect of visible light images is poor in smoke,night and other conditions.Because the infrared and visible images in the same scene have good complementary characteristics.Therefore,the effective fusion of the target information in the infrared image and the background detail information in the visible light image can improve the recognition of the target and obtain a more comprehensive and accurate description of the scene,which is of great significance for subsequent applications such as military operations,security monitoring and vehicle driving.Infrared and visible image fusion technology is one of the most widely studied and applied technologies in the field of multisensor image fusion.Aiming at the problems of low contrast,detail loss and poor definition in the process of infrared and visible image fusion,an image fusion method based on sub-window variance filter(SVF)multi-scale decomposition is proposed.According to the principle of the sub-window variance filter,the performance of the sub-window variance filter is compared with that of the classical edge-preserving filter by experiment,and it is concluded that the sub-window variance filter has better gradient and contrast retention characteristics.In order to realize the multi-scale decomposition of the source image,a multi-scale decomposition method based on the subwindow variance filter is proposed on the basis of the excellent characteristics of the subwindow variance filter,so that in the process of image decomposition,the problem of image contrast reduction can be suppressed as much as possible.The source image is decomposed into detail layer images containing high-frequency texture information at different spatial scales and base layer images containing only large-scale low-frequency contour and structure information.Different fusion rules are designed according to the characteristics of the base layer and detail layer images.The function of singular value in image processing is analyzed through experiments,and the basic image fusion rules based on singular value decomposition are designed.The pulse-coupled neural network model is introduced and applied to detail level image fusion with the help of image gradient value to improve the detail information in the fusion image.Finally,the fused base layer image and detail layer are reconstructed to obtain the final fused image.The experimental results show that the method preserves the details and salient target information in the source image well,and has good visual contrast and clarity.In order to further improve the efficiency of the image fusion algorithm,and to further preserve the high-frequency information of the visible light image and further integrate the significant thermal target information of the infrared image into the fused image,a hybrid multiscale decomposition method based on sub-window variance filtering and Gaussian filtering is proposed.,for pixel-level infrared and visible light image fusion.By studying the difference between the results of sub-window variance filtering and Gaussian filtering,an image decomposition method is established to further decompose the source image into different scales and different types of small-scale detail layers and edge layers,large-scale detail layers and edge layers,and base layer image.For the small-scale detail layer and the edge layer image,the "absolute value" is taken to be large for fusion.For the large-scale detail layer,edge layer and base layer images,the fusion weight map is constructed according to the edge information in the separated edge layer images,and the fused images of each layer are reconstructed to achieve the final fusion.The experimental results show that the fused image achieved by this method has clear edges between the infrared target and the background,rich visible light information,good subjective effects and objective evaluation results,and at the same time,the algorithm efficiency has been greatly improved.