Abstract: This paper improves the traditional Mixture of Dirichlet Process(MDP) nonparametric algorithm . The Markov Random Field(MRF) as spacial neighborhood is involved, and a new nonparametric image segmentation algorithm is constructed in this paper. Combined with MRF special neighborhood prior, the image segmentation results are smoothed and the cost of time is reduced. Comparative experiments in noisy natural images and magnetic resonance images show that the accuracy of improved algorithm is higher.

Key words: image segmentation, Bayesian method, clustering, Mixture of Dirichlet Process(MDP), Markov Random Field(MRF)

摘要： 对传统的Dirichlet过程混合(MDP)非参数算法进行改进，提出一种新的MDP非参数图像分割算法。引入马尔可夫随机场(MRF)空间领域关系，并将其作为空间先验约束条件对图像后验概率加以约束。该算法能够光滑图像中的边缘部分、控制分类数并加快收敛速度。实验结果表明，与传统算法相比，该算法的分割准确度较高。

关键词: 图像分割, 贝叶斯方法, 聚类, Dirichlet过程混合, 马尔可夫随机场

CLC Number: 

• TP391