Research On Image Fusion Method Based On Visual Salient Feature And Local Structure Consistency

Image fusion combines multiple images of the same scene into a new image with more comprehensive information and richer content.It is an effective way for multiple sensors to cooperate with each other in practical application,which can obviously improve the deficiency of single sensor imaging and enhance the stability and reliability of the system.At present,it has high application and research value in military,video surveillance,digital photography,medical diagnosis and so on.However,due to the particularity of images collected by different types of sensors and the complexity of information extraction in images,image fusion technology has not achieved the desired effect.Image fusion technology involves two key factors,image representation and fusion strategy design.In the case of a certain image representation method,the design of the fusion strategy is particularly important.In this paper,the problems existing in the method of multi-scale decomposition are studied in depth,and the frontier progress of image visual salient features and local structure consistency processing is combined.The main work includes:(1)For the fusion of infrared image and visible image,the infrared target is not salient and the background details are not rich.This paper proposes an infrared and visible image fusion algorithm based on multiple visual features.At the same time,local and global salient features of the source image are taken into account.The fusion is implemented under the non-subsampled contourlet transform(NSCT)based image fusion framework.The sub-band coefficients are fused with the weight map which is constructed by considering both visual saliency uniqueness and taskoriented objectness,and refined by spatial consistency with guide filter.The new fusion strategy highlights the prominent infrared objects of source image,and the visual background is more natural with more detailed information,thus effectively improving the quality of the fusion image.(2)In view of the seam effect in multi-sensor image fusion,this paper proposes an image fusion method based on structural saliency and content adaptive consistency verification.The algorithm uses the structural gradient features and visual saliency feature of the source image.Comprehensive features are obtained by controlling the weight proportion of features to guide different types of image fusion,which can effectively avoid incomplete extraction of source image information by single feature.In multi-focus image,infrared and visible image,medical image fusion experiments,the subjective and objective evaluation results show the algorithm's good accuracy and universality.