| The incidence of infant brain disease is becoming higher and higher.At present,the research of computer aided diagnosis of infant brain disease are still on the stage of theoretical research.In the use of computer applications with a variety of image analysis method of brain diseases assistive technology,the most critical and important issue is extraction different brain tissues in the infant brain MR images,This thesis propose three brain tissue segmentation and extraction algorithms for different characteristics of infant brain MR images.(1)Markov random field image segmentation algorithm based on Markov kernel function.This thesis use image block instead of single pixel and calculate pixel weight values in image block by Gaussian kernel function,and then this algorithm use KL distance to measure pixels the prior probability similarity in the image blocks and introduce a penalty term in the mode,finally,the maximum a posteriori probability solve the model.The experimental results show that this algorithm can effectively segment infant brain MR images with serious noises.(2)Markov random field image segmentation algorithm based on SVM.Under MRF model framework,this algorithm have a the unsupervised learning of image using SVM to obtain image prior probability,and then,this algorithm creat a model for imge using Gauss mixture model to obtain image probability distribution.finally,using maximum a posteriori probability of the model solve the model.The experimental results show that the algorithm can effectively segment infant brain MR images.(3)Weighted rough set fuzzy clustering image segmentation algorithm.this algorithm add weighted fuzzy factor and the Gauss radial basis kernel function into rough set fuzzy clustering model.and this algorithm depict the fuzzy problem of the boundary between the brain tissue by rough sets and fuzzy sets.Experiments show that this algorithm can effectively segment infant brain MR images affected by the noise and PV effect. |
PDF Full Text Request