The Study Of Improved PCNN In Multimodal Medical Image Fusion

The purpose of medical image fusion is to solve the problem that due to the limitation of imaging equipment,the diagnosis of some diseases requires the combination of several medical images in different modes,including from image fusion and general information fusion to a technique that reflects medical problems through images of human organs and cells.Digital image fusion is a comprehensive information processing of multi-source images,which provides a composite image containing the key information of the source image.This information plays a crucial role in better locating different organs and lesions to obtain a more accurate,comprehensive and reliable description of specific areas or targets for subsequent image analysis and understanding by doctors.Multimodal medical image fusion is widely used in the diagnosis and treatment planning of various diseases.Multimodal medical image fusion is widely used in the diagnosis and treatment planning of various diseases.Multimodal medical image fusion is widely used in the diagnosis and treatment planning of various diseases.At present,the research on medical image fusion mainly includes weighted average,multi-scale decomposition and neural network and so on.Aiming at the problems of the existing medical image fusion algorithm,this paper focuses on the improvement of the algorithm of Pulse-coupled Neural Network(PCNN)based on the Non-Subsampled Shearlet Transform(NSST)domain,and combine with other two methods.One is with Compressed Sensing(CS)the other is with Phase Congruency(PC).The main work and innovations include the following:(1)In allusion to the issue of the traditional PCNN fusion algorithm,in this paper,a simplified PCNN model is proposed to improve the external excitation input and link strength,so as to reduce the number of parameters,reduce the computational complexity,and improve its performance.In addition,solve the problem that is difficult to determine the number of iterations in the traditional model.(2)A medical image fusion algorithm based on NSST combining improved PCNN with compressed sensing is proposed.Firstly,decompose the source images into high frequency coefficients and low frequency coefficients by NSST.Then,the proposed algorithm fuses high frequency subband coefficients,and the observed values obtained from the source image compression and sampling are used as the feedback input of PCNN.The low-frequency subband coefficients are directly treated by regional feature weighted fusion rule.Finally,the final fusion image is obtained by using the transformation of inverse NSST.The experimental results show that the proposed algorithm achieves better results in both subjective and objective evaluation indexes,and the processing efficiency is improved.(3)In order to better retain the edge information of the source images,a high frequency coefficient algorithm combining PC with PCNN is proposed.The PC value is used as the external excitation input of PCNN.The low-frequency subband coefficients adopt an image fusion rule of local Laplacian energy.The measurement of structured information is realized using weighted local energy and weighted sum of eightneighborhood based modified Laplacian based on improved Laplacian energy.Compared with the traditional image fusion algorithms in the same field,the experimental results show that compared with the other three algorithms,the fusion image obtained by using this algorithm has better fusion quality,the edge information is kept intact,and the fusion time is reduced.The results of subjective and objective evaluation are consistent.