Research On Image Fusion Methods Based On Adaptive Decomposition

Image fusion is an important field of digital image processing and computer vision,which has received a lot of attention from researchers at home and abroad.As an important class of image fusion methods,transform domain methods have good detail performance and can better fuse different types of images.However,classical transform domain methods such as wavelet transform and discrete cosine transform usually use fixed basis functions without adaptive decomposition according to the nature of the image itself,which affects the quality of image fusion.To this end,this paper investigates image fusion methods for adaptive decomposition and designs two different image fusion methods considering the respective characteristics of multi-focus images and infrared and visible images,and the main work includes:(1)In order to improve the quality of multi-focus image fusion,this paper proposes a multifocus image fusion method based on BEMD(Bi-dimensional empirical mode decomposition)with joint bilateral filtering.On the basis of this method,the first embedded mode function’s variance maximum is calculated as the initial weight mapping,and then the joint bilateral filter is used to smooth it to obtain an improved weight mapping,and then the decomposed embedded mode function and the residual are fused with it;finally,the fused embedded mode is firstly decomposed into several embedded mode functions and a residual by BEMD,which can The method firstly decomposes the source image into several embedded modal functions and a residual by BEMD,which can effectively separate the feature information of different scales of the source image.The functions and the residuals are reconstructed to obtain the final fused image.The experimental results show that the fused images obtained by this method not only have high visualization quality,but also have better objective evaluation results in terms of mutual information,structure-based similarity and other measures compared with some existing image fusion methods,which can better improve the problem of artifacts at the focus boundary of multi-focus image fusion.(2)In order to improve the quality on the fusion of infrared and visible images,this paper proposes a novel method based on structure and texture-aware Retinex(STAR).It firstly decomposes the source images into reflection and illumination components according to the STAR model.This decomposition can not only separate the texture and structure of the source images accurately,but also extract the detail features of the visible images with low luminance well.And then,it merges the reflection component with the help of a weight map,which is constructed by using the second-order gradient of the source images as the input.Besides,it merges the illumination component with the help of a gamma function,which can make the fused image have more brightness information.Finally,it reconstructs the fused reflection and illumination components to obtain the final fusion image.A large number of experimental results show that the fused image obtained by this paper not only has high visualization quality,but also achieves better objective evaluation results in terms of mutual information,nonlinear correlation information entropy and image phase consistency measure compared with various methods proposed in recent years,which can better solve the problems of missing detail features and low contrast in the fusion of infrared and visible images.