Research On Research On Multimodal Brain Image Fusion Method Based On Riesz Wavelet Transform

With the development of biomedical engineering technology and computer technology,a variety of medical advanced imaging equipment with different functions have emerged,which provides a variety of modal medical images for clinical diagnosis and treatment.Nevertheless,a single-modal medical image limits to reflect the comprehensive information of morphological,structural and the function of ones tissues and organs,which is not conducive to the diagnosis of doctors and cannot meet the increasing clinical diagnosis needs.If the complementary information of different modal images is fully exploited,it will help the doctor understand the lesion information more comprehensively and then locate the lesion more accurately.Taking multi-modal medical images including MRI?CT?PET and SPECT as the research objects,thesis optimizes the image fusion quality mainly from the image decomposition and reconstruction.In view of the fact that the Riesz wavelet transform and its high-order generalization have the characteristics of preserving frame boundaries when mapping,and have a better representation of high-order functions,it is used as the tool for image decomposition and reconstruction.Aiming at the phenomenon that artifacts and distortions appearing in medical image fusion,thesis makes full use of the advantages of Frei-Chen operator in extracting image points,lines,space,edges and other features Based on it,a fusion rule combining salient features is constructed which is used for fusing the low frequency sub-band.As for the high frequency sub-band fusion,a gray-level saliency fusion strategy is adopted directly.Thus,a fusion algorithm based on Riesz wavelet transform and edge detection is proposed.In order to extract the global features of the image further and retain more image detail effectively,thesis proposes an image fusion frame based on the attention mechanism and improved PCNN.Specifically,the Frei-Chen operator is alternately used as a convolution kernel,and an image fusion attention mechanism is constructed based on the convolution result,which is used for the fusion of low frequency sub-band.Considering that the high frequency sub-band contains rich texture,brightness and color information,and the superiority of the PCNN model in processing high frequency signal,thesis fuses the high frequency sub-band coefficients by the PCNN model after optimized,that is,for the case that there are so many parameters and difficult to set in PCNN model,a parameter adaptive simplified model(PA-PCNN)is proposed,which the incentive is input by improved spatial frequency.The results show that the features can well represented based on the Riesz wavelet transform,and compared with the classic fusion methods in other literatures,the fusion frame based on the improved PA-PCNN combined with the attention mechanism achieves a better performance in vision,which has advantages in objective indicators.The research results provide a good choice for medical image fusion.