Research On Image Fusion Method Based On Sparse Representation Of Structure Group At Multi-scale

With the rapid development of artificial intelligence and information technology,image fusion is becoming more and more important.In the image fusion technology,people begin to decompose the source image by multi-scale method,so that the image fusion can be carried out in the discrete transform domain,and the multi-scale wavelet transform has achieved remarkable results in image fusion.In this paper,the wavelet transform image fusion method with multi-scale property and the image fusion method based on sparse representation of structure group are analyzed and studied.In view of the problems existing in wavelet transform image fusion method,this paper proposes an image fusion method based on structure similarity fusion rule.Then,according to the development and application of structure group sparse representation image fusion,an improved dual tree complex wavelet transform(DT-CWT)and structure group sparse representation image fusion method is proposed.The details are as follows:Firstly,this paper analyzes the background and significance of image fusion theory,and understands the status and development trend of image fusion theory at home and abroad.Next,this paper introduces the basic theory of image fusion and traditional fusion methods,and also introduces the evaluation criteria of image fusion quality.In the analysis and research of multi-scale transform theory,this paper analyzes the commonly used wavelet transform,and finds that the disadvantage of wavelet transform is that it does not have translation invariance and effective direction selectivity.Aiming at the shortcomings of wavelet transform,this paper proposes an improved image fusion method based on DT-CWT transform.This method effectively selects the components generated by DT-CWT based on structure similarity(SSIM),which solves the problems of wavelet transform and improves the image fusion effect.The experimental results and objective evaluation criteria show that the improved DT-CWT transform improves the image fusion effect.Through literature review,this paper analyzes the development process of sparse representation method and its application in image fusion.It is found that the traditional sparse representation has two problems: the difficulty of dictionary learning and calculation and the block correlation of sparse representation image.Aiming at the two problems existing in the traditional sparse representation,this paper analyzes the literature on the sparse representation of structural group,and finds that the sparse representation method of structural group can solve the above problems.At the same time,this paper introduces the structure group theory and the construction of the structure group sparse model.According to the advantages of sparse representation of image structure group and the advantages of DT-CWT transform,a multi-scale structure group image fusion method based on DT-CWT transform is proposed.This paper introduces the fusion process of this method,including the improved fusion rules of low and high frequency components of source image.Finally,the simulation results of the commonly used image fusion methods are compared with the simulation results of this method.It can be concluded that the proposed method can effectively improve the image fusion effect,improve the image clarity and retain the image details.