Medical Image Fusion Based On Improved PCNN And Sparse Representation In Nonsubsampled Shearlet Domain

With the continuous development of medical imaging technology,there are many advanced medical imaging equipment with a variety of different functions for clinical diagnosis and treatment.However,for the same human organ tissue,the single modal medical images can only reflect the limited structure,morphology and functions of the information,and not meet the clinical diagnosis and treatment needs.Medical image fusion technology aims to make full use of the useful and complementary information of different Multi-modal images to obtain more comprehensive and accurate information for helping the doctors improve the diagnostic rate.Taking different Multi-modal medical images for research objects,the image fusion is analyzed from the perspective of transform domain and fusion algorithms.An medical image fusion algorithm based on improved Pulse Coupled Neural Network(PCNN)and sparse representation in Nonsubsampled Shearlet domain is proposed.Due to the low frequency coefficient produced by the decomposition of the Nonsubsampled Shearlet Transform,the sparse representation is introduced to the low frequency image fusion.In order to preserve the more details of the image,the method of weighted sum of energy and variance is used to fuse the sparse coefficients.The sparsity of high frequency coefficients is good,and there is strong correlation between each pixel of frequency coefficients.For improving the fusion accuracy,PCNN is introduced in high frequency image fusion.Pointing at the poor fusion effect of the single pixel input PCNN,the Sum of the energy of Modified Laplacian is introduced as the PCNN input.Pointing at the limits of PCNN adaptability of fixed value as the link strength,the gradient energy of eight direction Sobel operator is introduced as the PCNN link strength for adjusting the PCNN link strength adaptively.The high-frequency fusion coefficients are selected by the sum of ignition frequency maximum.Taking some fusion experiments for the different Multi-modal medical images,the results show that using preferentially sparse representation on the low-frequency images,and using the improved PCNN fusion method on the high-frequency images are clearer than thoseusing the single fusion algorithms,and the fusion effect is better,and the proposed algorithm achieves good results both in the subjective and objective criteria.