Research On Fusion And Enhancement Methods Of Infrared And Visible Light Image Guided By Visual Saliency Features

Infrared and visible light image fusion is an important branch of multi-sensor information fusion,which is widely used in military,security,video surveillance and other fields.In view of some of the problems existing in the fusion of infrared and visible light images,researchers have proposed a series of improved algorithms,and the fusion results have achieved certain results.However,when the imaging scene is complex and changeable,the fusion image may have artifacts,low background contrast,target may be lost,salient features are not prominent,and visible light information is easily lost when the light intensity is low.Therefore,this paper proposes the following two fusion algorithms for the above-mentioned problems.The specific work is as follows:1.Aiming at the possible artifacts and ambiguities in the results of infrared and visible light image fusion,an infrared and visible light image fusion algorithm based on visual salient features and Poisson reconstruction is proposed.The algorithm realizes the fusion of image information in the gradient domain,effectively avoiding the phenomenon of false edges due to the inconsistency of local brightness due to the fusion results of multi-scale decomposition and sparse representation methods.In addition,the algorithm takes into account the local structure and visual salient features of the image,which is sufficient The texture information of the visible light image is retained,and the visually significant target information of the infrared image is highlighted.The fused image is clear and does not produce artifacts and noise.2.Aiming at the problem that the detailed information of visible light image is not prominent in low light environment,which leads to the loss of information in the fusion result,an infrared and visible light image fusion and enhancement algorithm based on visual salient features and roll guide filtering is proposed.The algorithm first enhances the low-light visible light image based on the Retinex theory,and successfully highlights its detailed texture information.Then,a rolling guidance filter(RGF)is used to decompose the enhanced visible light image and the source infrared image to obtain the respective detail layer and base layer.Finally,the fusion weight map is designed using the visual salient features of the source image,and the detail layer and the base layer are weighted and fused separately to obtain a natural and clear fusion image.The algorithm solves the problem that the detailed information of the fusion result is not prominent due to the low illumination intensity of the visible light image,the fusion result does not appear to be lost,and the target feature in the source infrared image is highlighted.