多分量混合高斯密度的一种参数估计算法

AN ALGORITHM FOR ESTIMATING THE PARAMETERS OF A MULTICOMPONENT MIXTURE OF GAUSSIAN DISTRIBUTIONS

根据有噪图像数据处理的需要,提出了一种分级矩估计与最大似然估计的综合算法来估计多分量混合高斯密度函数的参数.该算法首先利用分级矩估计法求出混合密度函数的初始参数估计值,然后利用最大似然估计法进行迭代运算,以求得这些参数的最终估计值.这种算法可以对具有2~m 或3×2~m(m 为正整数)个高斯分量的混合密度函数进行参数估值,给出了4,6,8分量混合高斯密度函数的参数估值实例.

A HMML algorithm for estimating the parameters of a mul- ticomponent mixture of univariate Gaussian distributions which have equal variances is presented.It is a combination of a hierarchical moment estima- tor (HM) and a maximum likelihood estimator (ML).The HM provides ini- tial values of the parameters for the ML.It can work for Gaussian mix- tures with 2~m or 3×2~m(m is a positive integer)components,and is applicable to estimation of noisy image data.Numerical examples are presented for four-,six-,and eight-component mixtures.