Research On Medical Image Fusion Algorithm Based On Fast Finite Shearlet Transform

Medical image fusion is the fusion of two or more medical images of the same part of the human body with different imaging principles,using fusion technology to turn into an image with comprehensive information to achieve a more comprehensive,accurate and detailed description of the human body.This has important significance and value for clinical medical diagnosis and treatment.In recent years,medical image fusion has gradually become a research hotspot.This paper addresses the problem that wavelet transform is easy to cause loss of detailed information,and the information will interfere with each other during image fusion,resulting in poor overall performance.The research focuses on medical image fusion algorithms in a multi-scale framework.The main research contents are as follows:(1)Aiming at the problems of poor comprehensive performance caused by the loss of detailed information caused by wavelet transform and mutual interference of information during image fusion,this paper proposes a medical image fusion algorithm based on FFST(Fast Finite Shearlet Transform)and improved PCNN(Pulse Coupled Neural Network).Use FFST to decompose the source image to obtain low-frequency and high-frequency coefficients.In order to overcome the shortcomings of low-frequency coefficients that are not sparse enough,use the sparse representation fusion method to fuse low-frequency coefficients without disturbing the details of high-frequency information.The high-frequency coefficients are fused using the improved PCNN,and the inverse FFST transform is used to obtain the fused image.The experimental analysis of multiple groups of different types of medical images shows that the fusion algorithm proposed in this paper has better performance,the fusion result better retains the source image information,and is more subjective to the human eye's observation.The performance is better than the comparison algorithm.(2)When FFST is used to decompose the source image,in order to make the fusion image more consistent with the human visual effect,an improved PCNN model is used to optimize the fusion algorithm for high-frequency coefficients.The pixel itself is used as the neuron's feedback input to stimulate each neuron,and SF(Spatial Frequency)and EOL(Energy of Laplacian)are selected as the neuron's link strength value,and the obtainedcumulative output is adaptively processed to obtain a weighting function to obtain a new ignition map to fuse high-frequency coefficients to achieve a better fusion effect.(3)In the FFST method,there are problems that low-frequency information is easy to lose and the subjective effect of details is not clear enough.In view of the strong information extraction capability of convolutional neural networks,this paper proposes a medical image fusion algorithm that combines FFST and convolutional neural networks in order to obtain more comprehensive information,detailed information such as edges and textures are more clearly fused images.This algorithm uses a convolutional neural network model to fuse high-frequency coefficients on the basis of FFST.Multiple sets of medical image experiments show that the fusion result obtained by the algorithm in this paper has better subjective visual effects and richer detailed information.