马尔可夫化的多尺度FCM在影像分割中的应用

Multi-resolution fuzzy C-means clustering based on MRF for image segmentation

为了同时处理影像分割问题中的随机性与模糊性,提出了一种多尺度(MR,multi-resolu-tion,马尔可夫随机场(MRF,markov random field)模型下的模糊C均值(FCM,fuzzy C-means)聚类分割算法(MR-MRF-FCM)。利用FCM算法能够处理影像模糊性的优点、MRF模型描述空间关系的长处以及小波的多尺度分析的优点,先对影像进行多尺度小波分解,并对小波系数建立MRF,进而用MR-MRF中的条件概率矩阵代替FCM算法的隶属度矩阵。实验结果从视觉效果和定量指标两方面表明,本文方法优于经典的MRF、多尺度MRF、FCM和核FCM等方法。

In order to simultaneously deal with both the stochastic nature and the fuzzy nature of images,a multi-resolution（MR） fuzzy clustering algorithm based on Markov random field（MRF） is proposed.The algorithm can combine both the advantages of fuzzy nature of the fuzzy clustering and the spatial description ability of the MRF model.The algorithm obtains the multi-resolution expression by decomposed wavelet.And the MRF of the wavelet coefficients is established,and then the conditional probability matrix in multi-scale MRF is used to replace the membership matrix in the fuzzy C-means clustering algorithm.Three experiments have verified that the proposed algorithm outperforms the classical MRF,multiscale MRF,FCM and kernel FCM based algorithms in terms of both visual quality and quantitative indices.