摘要
经典非参数Dirichlet混合过程模型图像分割算法具备在未知类数情况下实现图像自动分割的特点,但是由于其计算速度较慢,限制了该方法在临床上的实时应用.本文在经典非参数模型基础上进行改进,该算法首先将图像进行各项异性扩散滤波平滑,然后将马尔科夫随机场空间约束作为Dirichlet混合过程模型的先验进行分割计算.文中使用新算法对15例脑肿瘤磁共振图像进行分割实验,结果显示新算法能更有效控制收敛时图像分割类数,并且在图像分割的精度和计算速度等特性方面都明显优于经典的Dirichlet混合过程模型分割算法.
Based on nonparametric Dirichlet process mixture models, a novel automatic image segmentation method is proposed for brain tumor magnetic resonance images. In this paper all the brain tumor magnetic resonance images are smoothed by anisotropic diffusion firstly, then the images were segmented by the Dirichlet process mixture models combined with the Markov random field special restriction prior. The new algorithm is applied to segment the brain tumor magnetic resonance images, and the experiment results show that the class numbers can be controlled better in the new algorithm and the properties, such as accuracy and computing speed are significantly greater than the classical Dirichlet process mixture model segmentation.
出处
《小型微型计算机系统》
CSCD
北大核心
2013年第5期1181-1183,共3页
Journal of Chinese Computer Systems
基金
国家"九七三"重点基础研究发展计划项目(2010CB732500)资助
国家自然科学基金重点项目(30730036)资助
广州市属高校科研计划一般项目(2012A072)资助
