Edge-preserve Filter And Sparse Representation Fusion Method Based On Multi-model Medical Image

Medical image fusion is to preserve more important information in multiple medical images of different modalities,so that a method is used to synthesize input multiple medical images into one output image.The purpose of fusion is to provide more accurate auxiliary information for clinical diagnosis and treatment.This thesis focuses on multimodal sensor medical image fusion method based on edge-preserving filter and sparse representation.The basic framework of the study includes three parts: decomposition and reconstruction of medical images,image fusion rules,and image evaluation indicators.This thesis focuses on two aspects of the decomposition and reconstruction of medical images and image fusion rules,meanwhile proposes two new multi-scale medical image fusion methods.For the traditional medical image fusion methods,low contrast information and a small amount of detail information in the fused image.This thesis proposes a new image fusion method based on mutual-structure for joint filtering and sparse representation.Firstly,the source images are decomposed into a series of detail images and smooth image through the mutual-structure for joint filtering.Secondly,sparse representation is adopted to fuse coarse images and then local contrast is applied for fusing detail images.Finally,the fused image is reconstructed by the addition of the fused coarse image and the fused detail images.By experimental results,the proposed method shows the better performance on preserving detail information and contrast information in the views of subjective and objective evaluations.Aiming at the problem of loss of image detail information and color distortion in the traditional fusion methods,this thesis proposes an efficient image fusion algorithm.The fusion image is constructed by combining parallel features on multi-scale guide image filter model.Firstly,the input images are decomposed into a series of coarse and detailed images at different scales by guide image filter.Secondly,the parallel features are extracted from the decomposed coarse and detailed images to get the saliency maps.The saliency weighted maps of coarse images aim to preserve the structural information using color tensor.Meanwhile,the saliency weighted maps of detailed images work for extracting the detailed and luminance information by context-aware operator.Finally,the fused image is reconstructed by the fused coarse image and the fused detailed image.The fused coarse image is obtained by sparse representation.Experimental results demonstrate that the proposed algorithm shows the better performance among the other fusion methods in the domain of MRI-PET,MRI-SPECT and Gene image fusion.