Multi-modal Medical Image Fusion Using Improved Adaptive Sparse Representation

In recent years,adaptive sparse representation algorithm has been a hot topic in the research of multi-modal medical image fusion.Adaptive sparse representation is an improved algorithm for sparse representation by adaptively selecting dictionary.In this algorithm,image fusion is achieved by sampling image blocks and constructing multiple dictionaries according to the gradient direction of the image blocks,and then using these dictionaries to sparse represent images.This algorithm can denoise the image while image fusion,but it has some defects in maintaining the edge information,which leads to the loss of edge details.Therefore,this thesis mainly improves the adaptive sparse representation fusion algorithm and does the following work:(1)The development history and practical value of multi-modal medical image fusion are studied,and image fusion is introduced according to different fusion levels and fusion methods.Next,it masters the research status of this field in the past ten years at home and abroad.And then it introduces the fusion algorithm by classification,which laid the foundation for the following research.(2)Evaluating the fusion quality is an essential process of image fusion research.This thesis gives the evaluation methods from subjective and objective.In subjective evaluation,there are detail amplification method and difference method.In the objective evaluation index,according to whether there are standard reference images in the evaluation process,it is classified and summarized: no reference objective evaluation index and full reference objective evaluation index.The calculation process of various indicators is introduced in detail.Subsequently,the performance of different indicators is evaluated,and the effective indicators suitable for different types of image fusion are obtained.(3)For multi-modal medical images,multi-scale image fusion algorithms have better results.This thesis summarizes the flow chart of multi-scale fusion algorithm,which has two key steps: one is multi-scale transformation,the other is the selection of fusion rules.It introduces a variety of multi-scale transformation methods and different fusion rules,and gives the calculation process.Subsequently,several multi-scale algorithms are combined with sparse representation for simulation and evaluation,which provides ideas for improving adaptive sparse representation algorithm.(4)A multi-modal image fusion algorithm based on guided filter and adaptive sparse representation is proposed.Firstly the algorithm uses Gaussian filter to decompose the input image into detail layer and base layer.And then using guided filter to obtain the weight map of the base layer.Secondly the adaptive sparse representation is used to fuse the detail layer image,and weighted average fusion is performed according to the weight map of the base layer.Finally,the obtained detail layer and the base layer are added to obtain the final fusion result.(5)The second algorithm is based on Laplacian pyramid and adaptive sparse representation of multi-modal medical image fusion algorithm.The algorithm firstly decomposes the input image into four layers using the Laplacian pyramid.Secondly,adaptive sparse representation fusion is performed on each layer of the pyramid image.Finally,the inverse Laplace pyramid transform is used to obtain the fusion result.Through comparative experiments using three sets of CT and MRI images,the effectiveness of the two algorithms is verified and their advantages and disadvantages are comprehensively analyzed.It is concluded that the second algorithm is better than the first algorithm.