Research On Infrared And Visible Image Fusion Algorithms Based On Image Structure Preservation

Nowadays,various visual sensing technologies have emerged.But the concentration of technology also determines the need to fuse information from sources images of different imaging technologies.This will result in an image that conforms to the human visual system or facilitates computer recognition applications.This is the multi-source image fusion technology.The research focus of this paper is on the fusion algorithm of infrared and visible images.The main content and innovation points of this paper are summarized as follows:1.Guided by the human attention mechanism,the basic description of visual saliency detection is given.Firstly,the related algorithms in different periods are listed.And then two small experiments are carried out to deepen the understanding of the five classic algorithms.This part of the content mainly paves the way for the later research.2.Based on the analysis of the structural characteristics of infrared and visible images,a saliency detection and structure-preserving filtering algorithm is proposed.This algorithm is aimed at the prominent and simple situation of the target in the infrared image.Firstly,the salient target region is extracted from the infrared image.This region is classified into the foreground region.The final foreground fusion component is selected by the rule that the absolute value of the coefficient is large.Secondly,the background area is filtered by the structure-preserving principle to obtain the fusion weight.And then the background fusion component is further obtained.Finally,the fused image is obtained by combining the foreground fusion component and the background fusion component.Compared with other fusion algorithms,the algorithms proposed in this chapter have certain advantages in both visual perception and evaluation data.3.A fusion algorithm based on image structure and improved Haar wavelet transform is proposed.This algorithm deal with the case where the target is not prominent and the distribution is scattered.Firstly,through the gradient information in the two source images,the principle that the absolute value of the coefficient is large is used to obtain the corresponding structural features of the fused image.Then,using the relationship between Haar wavelet decomposition coefficient and gradient,the image gradient is transformed into the input of wavelet decomposition.The initial fused image is then obtained by inverse wavelet transform.Finally,the final fused image is obtained by adaptive histogram optimization.The experimental results of this chapter are compared with the existing algorithms.The effectiveness of the algorithm in this chapter is verified from both subjective and objective aspects.