Medical images have the characteristics of lower contrast and higher fuzziness in different organizations. This paper proposed a clustering algorithm, which was hased on multi-agent and gray-scale mathematical morphology operations. It combined agent technology and gray-scale mathematical morphology model in the algorithm. And it selected the structural element of the mathematical morphology as agent. It distributed each agents of different types randomly in the discrete spatial grid, and according to the type of its structural elements, it chose the appropriate operations with the given the neighborhood average operator to realize the image clustering in the same time control system driven. The algorithm did not require a priori knowledge and pre-processing operations, and it was not sensitive to the initial cluster points and need not prior input the number of clusters. This algorithm had distributed parallel computing capabilities and self-analysis capabilities. The experimental results show that the algorithm has the accuracy and reliability.%针对医学图像具有对比度较低,不同组织之间的模糊性较高的特点,给出一种基于多主体和数学形态学灰度形态运算的聚类算法.算法采用agent技术和多结构元素结合的模式,用结构元素做智能个体,每个不同类型的agents随机散布在离散空间格点上,在同时刻控制系统驱动下agents根据其自身结构元素的类型用给出的邻域平均算子自主选择作相应的运算进而实现图像聚类.算法无须先验知识和预处理操作,对初始聚类点不敏感,无须事先输入聚类簇数.算法具有分布式并行计算功能和自主分析能力.实验结果验证了该算法的可行性和可靠性.
展开▼
