一种基于高斯混合模型的改进EM算法研究  

Advanced EM algorithm based on Gaussian mixture model

作　　者：宋磊[1] 郑宝忠[1] 张莹[1] 闫丽[1] 卫宏[1] 刘建鹏[1] 李涛[1] 杨恒[1] 

机构地区：[1]西安应用光学研究所,陕西西安710065

出　　处：《应用光学》2013年第6期985-989,共5页Journal of Applied Optics

摘　　要：针对传统EM算法存在估计参数不具有最优性,以及在参数估计中需要人工参与等问题,提出一种基于高斯混合模型的改进EM算法。采用无人工参与的无监督思想,获取高斯混合模型对直方图拟合的最优参数组合。实验表明,该算法不仅能够快速地估计模型参量,而且能够给出最优参数,并在图像增强中使细节更明显,对比度更适中。In order to solve the disadvantages of traditional expectation maximization （EM） a gorithm which lacks parameters optimization and needs human operation when estimating pa- rameters, an improved EM algorithm based on Gaussian mixture model was proposed. The un- supervised theory was used to calculate optimal Gaussian mixture model parameters. The sub- jective and objective indices of experiments show that the algorithm can not only estimate pa- rameters quickly but also figure out the optimal parameters, making the detail more obvious and the contrast more moderate in image enhancement application.

关 键 词：EM算法 高斯混合模型 图像增强 

分 类 号：TN911.73[电子电信—通信与信息系统] TP301.6[电子电信—信息与通信工程]