Research On Multi-source Image Fusion Method Based On Sparse Representation

Due to the limitation of sensor imaging mechanism and material technology,it is difficult to describe all the information in the scene comprehensively,accurately and clearly with a single type of image data.Multi-source imaging sensors can provide different types of images in the same scene.The purpose of multi-source image fusion technology is to extract complementary information from multi-source channel image data as much as possible,and to generate a fused image with more abundant scene information.Nowadays,multi-source image fusion technology has been widely used in digital photography,video surveillance,medical diagnosis and remote sensing target detection.Aiming at the insufficient information preservation in multi-focus image fusion and infrared and visible image fusion,this paper combines with the advanced sparse representation models and studies the multi-source image fusion method based on sparse representation,which effectively improves the quality of fused image.The main research work is as follows:Concerning the problem that patch-based sparse coding suffers from the limited ability to preserve detail information,a fusion method based on joint convolutional analysis and synthesis sparse representation(JCAS)is proposed to better preserve the structure and details of the source images.Firstly,each source image is decomposed into base layer and detail layer by using the good image decomposition ability of JCAS.Secondly,the structural characteristics of the base layers are analyzed,and the fusion strategy based on Laplacian pyramid(LP)is used to fuse the base layers to obtain the fused image of base layer.Thirdly,the characteristics of detail layers are analyzed,and the fusion strategy based on region energy is used to achieve the effective fusion of detail layers.Finally,the fused base layer and the fused detail layer are superimposed to reconstruct the final fused image.The proposed method has better fusion performance than the existing fusion methods from the perspective of visual analysis and objective evaluation on common ten groups of gray images and twenty groups of Lytro color images.Focusing on the issue that low contrast and low brightness of fused image are caused by low rank sparse decomposition of single image,an infrared and visible image fusion method based on joint low-rank and sparse decomposition(JLRS)is proposed to improve the detail information and overall contrast of fused image.Firstly,infrared and visible images are jointly decomposed into three parts,i.e.,common low-rank component,specific low-rank components and specific sparse components,by using joint low-rank and sparse decomposition method.Secondly,the structural characteristics of the special low rank components are analyzed,and the fusion strategy based on non-sub sampled shearlet transform(NSST)is used to fuse special low rank components.Thirdly,the characteristics of specific sparse components are analyzed and the fusion strategy based on regional energy is used to achieve the effective fusion.Finally,the fused image is obtained by integrating the common low-rank component,the fused specific low-rank component and the fused specific sparse component.The proposed method is qualitatively and quantitatively compared with nine existing methods on three publicly test data sets,i.e.,Nato-camp?Bristol Eden Project and TNO.Experimental results demonstrate that the proposed method can provide better fused images in terms of visual quality and evaluation results than the existing fusion methods.