基于空间可信性聚类的图像分割算法

Image segmentation algorithm based on spatial credibilistic clustering

为解决噪声图像分割问题,提出了基于可信性测度并利用局部空间连续性的模糊聚类算法。采用可信性测度描述隶属度,去除了模糊c均值聚类中各像素对于所有类的隶属度之和为1的约束;并利用相异指标将局部空间信息引入聚类从而增强了抑制噪声的能力。隶属函数的参数可由数据集特点计算,削弱了参数选择的影响。提出预选准则以提高模糊聚类的稳定性。计算复杂度分析和实验验证了算法的可行性与实用性。结果表明:该算法在分割质量和效率等方面优于现有算法,适用于各种噪声图像的分割。

An algorithm based on a credibility measure and local spatial continuity was developed to improve noisy image segmentation. The constraint that memberships of a pixel across clusters must sum to 1 in fuzzy c-means is removed by use of the credibility measure. A dissimilarity index imposes the local spatial continuity to enhance the noise suppression. The membership function parameters can be computed based on the dataset characteristics so as to weaken the effect of the parameter selection. A pre-selection criterion is used to improve the stability of the fuzzy clustering. Analysis of computational complexity and numerical tests validate the feasibility and practicability. The results show that the algorithm has better quality and is more efficient than current algorithms for noisy image segmentation.