Research On Multi-band Images Synchronous Fusion Method

The purpose of image fusion is to extract significant features from multi-source images,so as to form the more accurate and reliable information perception of a scene.At present,it has been widely used in medical assistant diagnosis,automatic driving,UAV navigation and remote sensing target recognition,etc.With the development of detection technology,more and more detection methods will be developed,which brings opportunities and challenges to the development of image fusion.With the improvement of detection technology,more and more detection technology will be explored,which brings opportunities and challenges to the development of image fusion.Most existing methods are oriented towards two fusion objects,and leads them have to adopt the sequential fusion way.That means the intermediate fusion results will be repeatedly integrated with the unprocessed images until all images are involved.However,repeated fusion will obviously increase computational complexity,and defects of fusion effect will be gradually amplified.In addition,the fusion of multi-band images(far-infrared,near-infrared and visible images)has attracted much attention due to the different imaging principles and complex imaging background.To this end,this paper deeply studies the multi-band images synchronization fusion.The main research work and innovations are summarized as follows:(1)A multi-band images synchronous fusion algorithm based on Non-Subsampled Shearlet Transform(NSST)and fuzzy reasoning is proposed.When multi-scale fusion method is used to process multi-band images,local blur often occurs due to the irrationality of low-frequency fusion rules.To solve these problems,a multi-band image synchronization fusion algorithm based on NSST and fuzzy reasoning is proposed.Firstly,source images are decomposed into high-frequency and low-frequency images by NSST.Then,according to the characteristics of sub-images,the low-frequency information is intergrated by fuzzy logic reasoning,where the parameter setting of fuzzy reasoning system is studied emphatically,and some valuable conclusions are sumed up.For high-frequency information,the absolute value of pixel is used as the criterion of its fusion rules.Finally,on the basis of fused sub-images,the fusion result could be obtained by inverse NSST(INSST).Experiments show that fusion results could highlight these thermal targets,retain the main details from original images,and achieve good fusion effects.(2)A multi-band images synchronous fusion algorithm based on pixel brightness and spatial information enhancement is proposedAt present,most algorithms have failed to consider the imaging principle of source images,so the brightness information and spatial position relationship of these original salient features may be destroyed in the fusion process,which affects the visualization effect of fusion results.In this paper,the differences of imaging principle and characteristic in multi-band images are deeply explored.Specially,the near-infrared image is taken as the reference image,then these thermal targets,background and edge textures are enhanced by the far-infrared image and the visible image respectively,so as to achieve the purpose of fusion.Firstly,an optimization model of multi-band images synchronous fusion is established based on its imaging characteristics,and the specific meaning of the model is explained in detail.Then,the solving algorithm is constructed based on the alternating direction multiplier method(ADMM)according to the model properties.In addition,the selection of main parameters in the model is specially discussed.Finally,numrous experiments show that thses fusion results have been improved significantly in both visual effect and objective evaluation index.(3)A multi-band images synchronous fusion algorithm based on potential information association is proposedIn the process of multi-band images fusion,the potential relationship between source images and fusion result often fails to attract enough attention,which may directly weaken the relationship between input and output,and even cause some strange phenomena,such as blur,artifact and block effect,etc.Therefore,an optimization model about multi-band images synchronous fusion based on potential information association is proposed.Firstly,a fusion model based on representation learning is established by using the low-rank attribute of fusion image and its linear mapping relationship with the multi-band images.Secondly,the significant local information from source images is projected into the target space through the Laplace feature mapping matrix.Then,the solving method is constructed according to ADMM.Finally,the parameters configuration about the model is discussed in detail.Experiments show that the proposed algorithm could achieve the goal of multi-band images synchronous fusion well,and they are free from the interference of singular effects,such as local blur and artifact,etc.(4)A multi-band images synchronous fusion algorithm based on low-rank and sparse embedding is proposedIt is not difficult to find that there are usually many redundant features and singularities among multi-band images,both of which may disturb or even submerge these significant features of source images.To solve above problems,an optimization model about multi-band images synchronous fusion via low-rank and sparse embedding is proposed.Firstly,it is assumed that there exists a linear mapping relationship between source images and the fusion result,which helps to reduce the unpredictability of the result image.Secondly,the clean and distinctive eigenfeatures of multi-band images are extracted via low-rank and sparse embedding.Then,it gives the feature selection function to the mapping matrix,which could project the captured eigenfeatures into those fusion results.Finally,the optimization model for multi-band images synchronous fusion is constructed.In addition,the model solution is designed based on ADMM.Experimental results show that the algorithm is superior to many other algorithms in visual perception.In terms of objective indexes,those fusion results are also highly evaluated by comprehensive indicators.