Research On Medical Image Fusion Algorithm Based On Multi-scale Transformation

As an important image resource,medical image plays an important role in medical research.As the main way to obtain medical images,medical imaging equipment has its own advantages and limitations,one mode usually can not provide enough information for the entire medical diagnosis,thus affecting the doctor’s diagnosis of the disease.In order to solve this problem,the most common method is to fuse multiple feature maps together in some way to meet the comprehensive evaluation of the information contained in various images,and expect to achieve the purpose of one plus one greater than two.Therefore,the application of medical image fusion technology was born.It aims to integrate the information collected by various medical imaging equipment to achieve the purpose of comprehensively and intuitively reflecting the actual situation of the patient,and to provide powerful and accurate help for doctors to understand the patient’s condition.Multi-scale transformation has the characteristics of multi-resolution analysis and multi-direction,so it can better deal with complex and diverse medical images.Accordingly,this paper proposes two fusion algorithms for medical image processing based on multi-scale transformation.The main contents are as follows:(1)Aiming at the problem of loss of detailed feature information in the process of medical image fusion,a pulse-coupled neural network(PCNN)medical image fusion algorithm based on non-subsampling contourlet transform(NSCT)and discrete wavelet transform(DWT)is proposed.NSCT has a good effect in processing image texture edges,and DWT has outstanding advantages in preserving image detail information.Therefore,these two decomposition methods are combined to effectively preserve image texture edges and detail information.Firstly,the source image is decomposed by NSCT transform to obtain the corresponding high and low frequency sub-bands.Since the low-frequency sub-band contains the main energy information of the source image,DWT is used to decompose the obtained low-frequency sub-band.Then,the PCNN is used to fuse the low frequency subbands,where the input items is the improved Laplacian energy sum.The fusion of high frequency subbands is realized by combining information entropy and matching degree.Finally,the multi-scale inverse transform is used to reconstruct the coefficients to obtain the fusion result.(2)Aiming at the complex and changeable characteristics of medical images,combined with the characteristics that non subsampled shearlet transform(NSST)can effectively capture image feature information,a local Laplace energy medical image fusion algorithm based on NSST is proposed.Firstly,the source image is decomposed by NSST and the corresponding high and low frequency subbands are obtained.Then,it is proposed to fuse the low frequency subbands based on the local Laplace energy extracted by the direction,and combine the regional energy and the average gradient to fuse the high frequency subbands.Finally,image reconstruction is performed using inverse NSST transform to obtain the fusion result.(3)In order to verify the effectiveness and feasibility of the two algorithms,this paper conducts several sets of comparative experiments in combination with the actual situation,and selects relevant algorithms for comparison operations.The experimental results are evaluated and analyzed from subjective and objective aspects.Firstly,from a subjective point of view,both algorithms can effectively improve the performance of the fused image,retain a large amount of information in the fused image,and improve the visual perception of the image.Secondly,from an objective point of view,the two algorithms have generally better performances in common image quality evaluation indicators,and the fusion effect is better than other comparison algorithms.