邻域信息约束融合Student's t混合模型的医学图像分割  

Medical Image Segmentation Based on Student's t Mixture Model Integrating Neighborhood Information Constraint

作　　者：王燕贞[1] 陈志翔[2] 罗俊星 WANG Yan-zhen;CHEN Zhi-xiang;LUO Jun-xing(Zhangzhou Institute of Technology,Zhangzhou 363000,Fujian,China;Minnan Normal University,Zhangzhou 363000,Fujian,China)

机构地区：[1]漳州职业技术学院,福建漳州363000 [2]闽南师范大学,福建漳州363000

出　　处：《山西师范大学学报（自然科学版）》2021年第1期59-66,共8页Journal of Shanxi Normal University(Natural Science Edition)

基　　金：福建省中青年教师教育科研项目(JAT191414);(JAT201274).

摘　　要：医学图像的高噪声以及偏移场的存在使得传统图像分割方法在分割此类图像时分割效果不理想.针对上述问题,提出一种邻域信息约束融合Student's t混合模型分割算法.首先,获取像素点的邻域信息进行滤波操作以达到降噪效果;其次,构建Student's t混合模型,通过噪声平滑因子融合邻域信息约束修正先验概率;最后,利用最大期望(EM)算法求解,获取最大后验概率实现图像分割.实验结果采用DICE指标验证了算法的有效性.Traditional image segmentation is unsatisfactory due to high noise and the bias field of medical images.A Segmentation algorithm based on Student's t mixture model integrating neighborhood information constraint is proposed to solve the above problems.First,the neighborhood information of the pixel point is obtained and filtered to achieve the noise reduction effect;secondly,the Student's t mixture model is constructed to amend the prior probability by integrating the neighborhood information constraint with the noise smoothing factor;finally,the maximum posterior probability is obtained to achieve the image segmentation by adopting Expectation Maximum(EM)algorithm.The validity of the algorithm is verified by DICE index.

关 键 词：医学图像 高噪声 邻域信息约束 Student's t混合模型 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]