Research On Brain MR Image Segmentation Algorithm Based On Fuzzy C-means Clustering

With the rapid development of information science and technology,the amount of information is increasing explosively.Images are an important way for humans to obtain information.In clinical medicine,magnetic resonance(MR)images have the advantages of high contrast,high resolution,and multiple orientations,which are widely used in various brain researches.In order to extract key information in the images effectively,image segmentation is an indispensable part in image processing.But,there are always partial volume effect,gray unevenness and noise in the MR images of the brain which make it a difficult work to complete image segmentation accurately.Image segmentation algorithm based on fuzzy C-means(FCM)is a kind of classical algorithm,which can describe the phenomenon of tissue boundary blurring in MR images well.Therefore it has been widely applied in MR images of the brain.However,the FCM algorithm also has defects such as being easily trapped in local optimum and sensitive to noise.This thesis focuses on the image segmentation algorithm and does the following work.(1)In this thesis,the fuzzy C-means(FCM)clustering segmentation algorithm is selected by analyzing various segmentation algorithms.Aiming at the defect that the algorithm is easy to fall into local optimum,this thesis introduces an improved quantum particle swarm optimization(IQPSO)algorithm and proposes a fuzzy c-means based on IQPSO(FCM_IQPSO)for the segmentation of brain MR images.The algorithm can optimize its iterative optimization process and improve the defect of trapping into local optimization.It is found that the segmentation effect of the FCM_IQPSO algorithm is better than other algorithms by comparing the results of the simulation test.(2)FCM ignores the neighborhood pixel correlation of brain MR images,and its nonlinear data processing ability is poor.To these problems,the spatial information of the image is introduced into the objective function of FCM in this thesis.On this basis,the algorithm is extended to the kernel space by the Gaussian Kernel function.Then,a kernel fuzzy C-means algorithm based on spatial correlation(KFCM_SC)is proposed.Finally,a comparative test is carried out in this thesis.The simulation results show that the KFCM_SC algorithm can achieve good segmentation effect and is robust to noise.