Abstract: This paper presents a simple and effective segmentation method for color image. All possible uniform regions in the color image are obtained by using the histogram threshold technique. That is, the main peak values is found in order to initialize and merge the regions and obtain the uniform regions marked by the corresponding  clustering centers. The adaptive K-means clustering algorithm is given to improve compactness of the formed uniform regions. The experimental results show that, compared with IAFHA（improved ant colony-Fuzzy C-means hybrid algorithm） method, the proposed method can receive the less segmentation regions, fast segmentation speed  and have certain robustness.

Key words: color image segmentation, histogram threshold, K-means clustering

摘要： 提出了一种简单有效的彩色图像分割方法。应用直方图阈值技术获得彩色图像中所有可能的均匀区域，即通过寻找主峰值，对区域作初始化和合并，获得由相应的聚类中心标注的均匀区域，提出自适应K均值算法以提高已经形成的均匀区域间的紧密度。实验结果表明，该方法与IAFHA方法比较，分割的区域数较少，且分割速度较快，具有一定鲁棒性。

关键词: 彩色图像分割, 直方图阈值, K-均值聚类