计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (30): 202-205.

• 图形、图像、模式识别 • 上一篇    下一篇

融合聚类和分级区域合并的彩色图像分割方法

刘 彬,王朝英,侯志强   

  1. 空军工程大学 电讯工程学院,西安 710077
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-10-21 发布日期:2011-10-21

Color image segmentation algorithm using clustering and hierarchical region merge

LIU Bin,WANG Zhaoying,HOU Zhiqiang   

  1. Telecommunication Engineering College,Air Force Engineering University,Xi’an 710077,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-10-21 Published:2011-10-21

摘要: 提出了一种融合聚类的分级区域合并彩色图像分割方法。为平滑图像且保持良好边缘,首先用均值偏移算法进行滤波,在此基础上运用改进的k均值聚类方法在颜色空间对图像进行聚类,形成图像的初始分割区域。融合颜色、空间和邻域信息度量区域的距离,对初始分割区域进行分级合并,直至满足停止区域合并的准则。利用形态学腐蚀与膨胀算法对区域边缘进行平滑。仿真结果表明,算法的分割结果符合人类主观视觉感知,具有良好的一致性。

关键词: 彩色图像分割, 均值偏移算法, k均值聚类, 区域合并, 边缘平滑

Abstract: A new algorithm of color image segmentation using clustering and hierarchical region merge is proposed.First of all,the mean shift algorithm is introduced to smooth the images while preserving the boundaries.The processed images are clustered by using the improved k-means clustering algorithm in the feature space,and then the initial regions are formed.Afterwards,the initial regions are hierarchically merged on the basis of the region distance defined by the color,space and adjacency information until the criteria of the merging termination is satisfied.The erosion and dilation operators are used to smooth the edges of segmented regions.Accoding to the simulation results,the color image segmentation results of the proposed approach are favorably consistent with the subjective visual perception of the human beings.

Key words: color image segmentation, mean shift algorithm, k-means clustering, region merge, edge smoothing