Research On Medical Image Fusion Methods Based On Multi-scale Transform And PCNN

With the development of imaging technology and imaging equipment,medical images of various modes have been invented,which greatly promotes the development of medical imaging technology.However,due to the different imaging principles and the physical limitations of imaging equipment,single mode medical images cannot fully reflect the information of a certain part or organ of the patient’s body,and various modes cannot replace each other.Therefore,it is of great significance to solve the problem of multimodal medical image centralization and redundancy removal,and the multimodal image fusion technology is also adapted to the times.Image fusion methods based on multi-scale analysis are research focus in recent years.Compared with the Wavelet transform and Contourlet transform,the Nonsubsampled Contourlet Transform has advantages such as direction features,multi-scale features and translation invariance.It is more suitable for image fusion.Pulse Coupled Neural Network conforms to the physiological mechanism of human visual system and has the characteristics of synchronous pulse release and global coupling.Therefore,this paper first proves the optimal filter combination of NSCT in the field of image fusion through experiments,and then makes an in-depth study on the fusion method based on PCNN in NSCT domain.The main research results and conclusions of this paper are as follows:(1)Aiming at the problems of traditional PCNN model,such as the inflexible parameter setting,the timeliness and the poor accuracy,an image fusion framework in NSCT domain is designed,whose key parameters can be adjusted adaptively according to image features.Experiments show that the algorithm based on this model can significantly improve the fusion quality of multimodal medical images.It can completely retain the valid information of the source image,and obtain the fusion images with higher contrast and better visual effects.(2)Aiming at the problems of volume effect and edge blur of MRI images,a fusion algorithm combining guided filter and PCNN is designed which is more suitable for CT and multiple weighted MRI images.The improved bootstrap filter is introduced into the fusion rule to highlight the image edge features and detect slight changes.Experiments show that this algorithm can preserve the bone edge of CT and soft tissue details of MRI completely and clearly.It can also reflect the subtle differences sensitively.The fused image has fine texture and high contrast.(3)Aiming at the problems of unclear outlines and details and low color contrast of color image fusion,a SFLA-PCNN optimization model based on the improved simplified PCNN is designed which can realize the full parameter adaptive adjustment of PCNN.The model makes use of the advantages of Shuffled Frog Leaping Algorithm such as easyto implement,fast convergence,few parameters and strong ability to search for results,and takes SFLA into the fusion framework for multimodal color medical images fusion.The fitness function is improved by combining information entropy with Rastrigin,which can take advantage of the inherent feature information of the source images.Experiment results show that the fusion image has a higher correlation with the source image information.The contrast of the color fusion image is obvious,which fully reflects the edge features.The information of the fusion image is helpful for clinical diagnosis.