The Research Of Medical Image Fusion Based On Sparse Representation

In order to obtain more accurate and richer medical image information,joint medical imaging equipment has been introduced.However,there are still many problems in the medical image fusion technology as the core part.Therefore,it is necessary to develop a fast,accurate,and stable medical image fusion system to meet the needs of doctors and researchers for higher dimensional medical information.Firstly,the medical image fusion algorithms in recent years based on spatial domain,transform domain and sparse representation are compared and analyzed in this dissertation.Image fusion algorithms based on sparse representation are the most efficient at present.However,they still have two most important problems:one is that the source image structure information cannot be taken into account in the algorithm,and the other is that the time complexity of the algorithm is too high.This dissertation proposes a medical image fusion algorithm based on sparse representation and NSCT.And combined with the‘local fusion rule',it successfully takes the structure information of the source images to the fused images.Comparing to the traditional sparse image based medical image fusion algorithm,it is improved by 16.5%,6.7%,11.1%,14.4%and 21.2%respectively on the performance parameters Q₀,Qe,Qw,Qab/f and MI,but it does not solve the problem of high time consumption.Then,in this dissertation,the influence of the size of the slider on the fusion result and the fusion speed is analyzed,and it is found that the quality of the fusion result and the fusion speed are best when the size of the slider is 6-9.Then,combined with"energy structure decision map",a medical image fusion algorithm based on feature extraction and sparse representation is proposed.Comparing to the traditional sparse image based medical image fusion algorithm,the time consumption of SR+SEM is reduced from 264.272s to 24.160s.Experiments show that the fusion results with high quality and the low time consumption of the fusion process can be guaranteed.Finally,this dissertation implements the SR+SEM software designing and system debugging on the TMS320C6678 EVM platform,and the corresponding hardware running results are still with high quality,and the average running time is only 1.131s.