中文摘要：

为改善传统模糊G均值聚类算法的抗噪性能，Krinidis和公茂果等提出像素局部邻域信息模糊C-均值聚类算法系列，但其存在聚类中心表达式与聚类目标函数不一致的问题。利用拉格朗日乘子将模糊局部信息聚类目标函数和隶属度约束条件相结合，构造无约束优化函数，利用函数极值存在的必要条件推导该聚类新的隶属度和聚类中心迭代表达式，设计一种核空间模糊局部信息C-均值聚类分割算法。人工合成图像和实际遥感图像分割测试结果表明，该算法明显优于现有模糊局部信息C-均值聚类分割法，针对复杂遥感图像能获得更好的分割效果。

英文摘要：

To improve the performance of traditional fuzzy C-means clustering algorithms in restraining noise, a series of local neigh- borhood pixels information fuzzy C-means clustering algorithms are put forward by Krinidis and Gong Maoguo et al. However, the iteration formulas of clustering center and objective function are not consistent, then the unconstrained optimization function was established by means of the Lagrange multiplier to combine fuzzy local information clustering objective function with mem- bership degree constraint conditions, and the new iteration formulas of fuzzy membership and clustering center were obtained using the necessary condition of the constrained extreme value of the function and rigorous mathematical analysis. Therefore, the kernel space fuzzy local information C-means clustering segmentation algorithms were presented. Results of synthetic images and remote sensing images segmentation tests show that the proposed image segmentation algorithms is superior to the existing local information fuzzy C-means clustering segmentation methods, especially for complex remote sensing image, it can obtain better segmentation effect.