Medical Image Fusion Algorithms Based On Multi-scale Transformation Of Feedback Mechanism

With the development of medical imaging technology,a large number of reliable single-modal medical images are provided for clinical diagnosis.But these images reflect different emphasis on a particular feature of organ.The information provided by a single-modal medical image is partial and incomplete.How to present different modal medical image information on the same image is a research topic in universities.The topic has been discussed in this paper,the core is research of multi-scale medical image fusion algorithm based on feedback mechanism.Through image fusion technology,the fusion of different modal medical images is achieved,so as to achieve the effect of image information enhancement,and get more comprehensive and high-resolution images.It provides a reliable basis for the diagnosis of the disease and the proposal of the diagnosis and treatment.The research content and innovation points are as follows:In this thesis,the imaging principles of medical images(such as CT,MRI,PET and so on)are analyzed,and their characteristics are summarized and summarized.Then the application of spatial domain algorithm,frequency domain algorithm,wavelet transform,multi-resolution geometric transformation in image field is analyzed.A comparative analysis of three sets of theories,such as spatial domain and frequency domain,wavelet transform and contourlet transform,contourlet transform and non-subsampled contourlet transform(NSCT),is presented.An adaptive multimodality medical image fusion algorithm based on NSCT transform is proposed.According to the different characteristics of high and low frequency sub-band image after decomposition,the following fusion methods are proposed: the low-pass band image reflects the basic information of the source image,and the method of combining the sub-block and the region energy is adopted;the image of the high frequency sub-band presents the details of the edge of the source image,and the improved sum of Modified Laplacian are used;At the same time,combining with the theory of feedback system,the genetic optimization algorithm is used to adjust the variable parameters in the fusion rules adaptively,and the optimal fusion coefficient is obtained.Finally,the coefficients are reconstructed to get the fused image.In order to overcome the disadvantages of non-subsampled contourlet transform,such as large amount of data,high computational complexity and poor real-time performance,a medical image fusion algorithm based on non-subsampled shear wave transform is proposed.Compared with NSCT,NSST computing is more efficient,more flexible and better image processing performance.In the design process of fusion rules,combined with pulse coupled neural network,imitation of human visual characteristics,the fusion results are more consistent with the human visual system.For low frequency sub-band images,the fusion method of regional ignition intensity is adopted,and the high frequency sub-band images are composed of visual sensitivity coefficient,region energy,improved gradient energy and the combination rule of the three,and the particle swarm optimization algorithm is combined to make the closed-loop system.The high and low frequency sub-band coefficients are fused,and the fused image is reconstructed by inverse NSST transform.Using two value and color images as source images,fusion image is obtained by applying the algorithm proposed in this paper and other scholars,and analyze the experimental results from subjective and objective two sides.The experimental results show that the fusion image information obtained by this algorithm is more abundant,the resolution is improved obviously,and the inheritance of detail information is better.From the analysis of image quality evaluation index,the some evaluation index of the fusion image obtained by this algorithm is the best,and the comprehensive evaluation index is good,which indicates that the algorithm can inherit source image information better,and pay more attention to detail integration.It is proved that the two algorithms proposed in the text are feasible and effective,and can get satisfactory fusion results and have practical application value.