Research On Application Of Variational Method In Image Fusion

Image fusion is an important module in the field of image processing,and it is a prerequisite work for many image processing links.Image fusion According to the source image acquisition device can be divided into homologous fusion and heterogeneous fusion.Based on the theory of functional and variational method,two kinds of image variational fusion models are proposed.The former fusion model is a branch of the fusion of homologous images:Pan-sharpening,the latter fusion model is a branch for heterogeneous image fusion:infrared and visible light image fusion.The two variational fusion models are similar in the basic idea:the corresponding energy functional is constructed based on the characteristics of the source image and the corresponding fusion type,and the fusion result is obtained by minimizing the energy function.In the numerical optimization method,this paper introduces the splitting Bregman iterative algorithm and the augmented Lagrangian multiplier method,which performs more efficiently and more stable when dealing with energy functional with L1 norm.The main contents of the thesis include:In the Pan-sharpening fusion direction,some of the current Pan-sharpening methods based on variational are mostly through the gradient descent method to minimize the energy functional to achieve fusion.However,the gradient descent method will slow down near the minimum.And if the variational model contains the L1 norm divisor,the gradient descent method is robust and computationally complicated.According to the L1 norm,the texture of the image is better than the L2 norm,and the Bregman splitting the Bregman to the function of the convergence speed of the L1 norm.On the basis of the existing variational model,the L1 norm is added to In the model,the energy functional cost function is constructed and the energy functional is minimized by the iterative minimization of the Bregman algorithm.Finally,the fusion result is verified by the worldview-2 and IKONOS satellite data.In the direction of infrared and visible image fusion,an energy functional model is proposed based on the existing research results of heterogeneous image fusion and the idea of homology image fusion.In the aspect of texture information retention,we refer to the idea of Pan-sharpening method,introduce L1 norm term,and add brightness information retention item to the functional model.In the minimization of energy functional optimization method,we choose the augmented Lagrangian multiplier method.Verify the fusion result with TNO_Image_Fusion_Dataset and OSU-Color-Thermal Database.