Liver CT Image Segmentation Algorithm Combining NLM Filtering And Improved FCM

In recent years,liver disease has gradually developed into a hidden danger affecting the health of our people.According to statistics,there are more than 300000 deaths from liver diseases in China every year,which has brought immeasurable disasters to countless families.With the development of medical imaging technology,computed tomography(CT)imaging technology is widely used in the diagnosis of liver diseases because of its convenient and rapid examination.Liver CT image is an important medium for doctors to check patients.How to quickly separate liver CT image from abdominal CT image is of great significance for clinical diagnosis and related research of patients.Therefore,it is necessary to propose an efficient algorithm for CT image denoising and segmentation.Firstly,this thesis studies the related technologies of CT image denoising,including image denoising algorithms based on spatial domain and frequency domain.For the noise of CT image is mainly additive noise and multiplicative random noise,the Non Local Means(NLM)filter is usually selected to study.However,when calculating the similarity weight,the filter only considers the image spatial domain information,uses the Euclidean distance to calculate the block similarity,and the algorithm complexity is high.For the problems existing in NLM filtering,this thesis adds pixel domain information to improve the similarity calculation,and uses the idea of integral graph to accelerate the algorithm in the calculation process.Then,the problem of liver CT image segmentation is studied.According to the fuzziness characteristics of medical images,the Fuzzy C-Means(FCM)algorithm is selected for research.It is found that FCM algorithm only considers the image gray information in the clustering process and needs to preset the initial parameters,resulting in poor edge detail processing of the segmented image.In view of this problem,an adaptive acquisition of initial parameters according to the image information is proposed,Add spatial information to improve the FCM algorithm.At the same time,combined with the region growth method,the liver image is finely segmented to obtain a clearer liver image.This thesis includes the following two innovations:(1)Aiming at the problem that NLM filtering only uses spatial domain information to process weighted information and has high algorithm complexity,the idea of bilateral filtering is used to improve NLM filtering,so that NLM filtering can combine the information of pixel domain and use integral graph to process the weight calculation process,so as to speed up the operation efficiency of the algorithm.In order to better process the noise image,the wavelet transform is used to divide the noise image into low-frequency component and high-frequency component,the improved NLM filter is used to process the low-frequency component,and the improved guidance filter is used to process the high-frequency component.Finally,the inverse wavelet transform is used to fuse the low-frequency component and high-frequency component to obtain the denoised image.(2)Aiming at the problem that FCM algorithm depends on initial parameters and only gray information for segmentation,LOF algorithm and histogram algorithm are used to select the appropriate number of points with large local outlier factor as the initial clustering center.The spatial information is obtained by using the relationship between the membership matrix and the a priori probability of Markov random field.The objective function and algorithm flow are improved.Finally,the improved region growth algorithm is used to obtain the complete liver image.In the experimental part,the algorithm proposed in this thesis is used to denoise and segment the liver CT image selected from SLIVER07 data set.The experimental results show that the denoising algorithm proposed in this thesis can effectively filter the mixed noise in liver CT,and the PSNR value and SSIM value are higher than the comparison algorithm;The segmentation algorithm proposed in this thesis performs better than the comparison algorithm in noisy and noiseless images,and retains the edge details of the image better.To sum up,this algorithm can effectively remove the noise in the liver CT image and completely segment the liver essence.