隐马尔可夫模型在对象定位中的应用与实现

以医学图像中脊柱的定位为应用背景,对经典Viterbi算法进行改进,根据全局最优搜索算法提出了可控的最优搜索算法,把隐马尔可夫模型(Hidden Markov model,HMM)应用于二维图像中线性结构对象的定位。以真实的人体脊柱X光照片为实验样本进行的定位实验,结果表明:HMM模型不需要对图像施加较多的限制,就可获得较成功的定位结果,同时引入了可控的最优搜索算法后,定位效果有了进一步的提高。还通过实验分析了可控制的最优搜索的两个控制参数对定位性能的影响,并提出了定位性能具有区间性的结论。

Application and Realization on Object Localization Based on Hidden Markov Model

A novel approach for localizing objects with the linearly two-dimensional structure is presented based on hidden Markov model(HMM).Based on the general best search localization algorithm a under-control best search localization algorithm is introduced by improving the classical Viterbi algorithm.Experiments on actual X-ray images of the spine show that no more constraints are applied to the image localization on HMM.The performance of the localization is improved by the under-control best search localization algorithm.The affects for two control parameters are compared on the localization performance.