3-D Medical Image Fusion Based On Tensor Low-rank Model

With the continuous development and innovation of medical imaging technology,medical imaging plays an important role in clinical diagnosis.Due to different imaging modes,medical images collected by different imaging devices present different modes,reflect different structural information of human body.However,the single mode medical images still can not describe the information of the lesion comprehensively.Image fusion is a kind of image processing technology that combines multi-modal images to produce a fused image.Image fusion can effectively improve the accuracy and efficiency of medical image aided diagnosis.Hecne,the research of multimodal medical image fusion has certain theoretical and practical value.Currently,there are few 3D image fusion methods,mainly for 2D images.But 2D image fusion methods can not effectively maintain the structural information between three-dimensional medical image slices.Therefore,this paper focuses on multi-modal three-dimensional medical image fusion based on tensor low rank model,aiming to promote the practical application of medical image fusion technology.The main controbutions and innovations of this paper are as follows: 1)A fusion method based on tensor robust principal component analysis(TRPCA)model is proposed.Tensor low rank decomposition structure can describe the similarity and special features between slices of 3D medical image.In this paper,trpca decomposition part of multimode 3D medical image is used as 3D feature,and ‘3D weighted local Laplacian energy' criterion and ‘absolute maximum' criterion are designed to fuse tensor low rank component and tensor sparse component respectively.Experimental results on synthetic and real 3D multimodal medical images demonstrate the effectiveness of the proposed method.2)A fusion method based on dual tensor low rank model(DTLR)is proposed.On the basis of TRPCA,the multi-component structure of the low rank part is further considered,and a double low rank model is constructed,including the low rank part of the foreground,the low rank part of the background and the sparse detail part.In addition,DTLR model is applied to 3D medical image fusion,and corresponding fusion rules are designed for each decomposition part.Compared with fusion method based on TRPCA,the experimental results verify that DTLR model can improve the fusion performance.3)A fusion method based on dual tensor robust principal component analysis(DTRPCA) model is proposed.Considering the relationship between the tensor low rank part and the source image,a DTRPCA model is proposed based on the low rank constraint.The decomposition part of DTRPCA is used as the feature of 3D medical image,and the corresponding fusion rules are designed for fusion.Experimental results on synthetic and real multimodal medical images demonstrate the superiority of the proposed fusion algorithm.