隶属度修正的空间核直觉模糊聚类苗族服饰图案分割

Spatial kernel intuitionistic fuzzy clustering algorithm for Miao costume pattern segmentation with modified membership

针对直觉模糊聚类算法未考虑空间邻域信息的问题,提出一种隶属度修正的空间核直觉模糊聚类苗族服饰图案分割算法。首先,在直觉模糊C均值(Intuitionistic Fuzzy C-Means,IFCM)算法的目标函数中加入隶属度约束惩罚项,降低算法的计算复杂度;其次,使用核距离替代欧式距离,重新计算像素点到聚类中心的距离,提升算法的鲁棒性;最后,考虑空间邻域信息,对邻域内的像素点赋予不同的权重来修正隶属度函数,从而完成图像分割。实验结果表明,在含噪声的苗族服饰图案数据集中,所提算法的划分系数和划分熵为95.49%和7.44%;在彩色苗族服饰图案数据集中,所提算法的划分系数和划分熵为97.91%和3.42%,均优于其他对比算法,且提高了核函数的运行效率。

Aiming at the problem that the intuitionistic fuzzy clustering algorithm does not consider the spatial neighborhood information,a spatial kernel intuitionistic fuzzy clustering algorithm for Miao costume patterns segmentation with modified membership was proposed.First,the membership constraint penalty term was added to the objective function of the intuitionistic fuzzy C-means(IFCM)algorithm to reduce the computational complexity of the algorithm.Second,the kernel distance was utilized as the proposed algorithm instead of the Euclidean distance to recalculate the distances from pixel points to cluster centers,which can obtain high robustness.Finally,the spatial neighborhood information was considered,different weights were assigned to the pixel points in the neighborhood to modify the membership function,and thus the final image segmentation results were achieved.The experimental results show that in the noisy Miao costume patterns dataset,the partition coefficient and partition entropy of the proposed algorithm are 95.49%and 7.44%.In the colorful Miao costume patterns dataset,the partition coefficient and partition entropy of the proposed algorithm are 97.91%and 3.42%,which are better than other comparison algorithms,and the operation efficiency of the kernel function is improved.