Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (18): 151-155.

Previous Articles     Next Articles

Multi-objective kernel clustering image segmentation based on watershed over-segmentation

ZHAO Feng, HAN Wenchao, HUI Fangchen   

  1. School of Telecommunications and Information Engineering, Xi’an University of Posts and Telecommunications, Xi’an 710061, China
  • Online:2015-09-15 Published:2015-10-13


赵  凤,韩文超,惠房臣   

  1. 西安邮电大学 通信与信息工程学院,西安 710061

Abstract: The traditional clustering-based image segmentation generally only utilizes the gray information of the image. In order to better use the region and edge information, a multi-objective fuzzy kernel clustering image segmentation algorithm based on watershed over-segmentation is proposed. The watershed method is used to over-segment the image and obtain some regions. Then the multi-objective fuzzy kernel clustering algorithm is used to cluster the representative points of the regions and the pixels on watershed. All the pixels of the image are labeled according to the clustering results of the region and the watershed to obtain the final image segmentation result. The experimental results show that the objective in the image can be more completely segmented from the background due to using the image region information.

Key words: multi-objective evolutionary algorithm, kernel clustering, image segmentation, watershed

摘要: 传统的聚类图像分割方法一般仅仅利用图像中的灰度信息。为了更好地利用图像中的区域和边缘信息,提出一种基于分水岭过分割的多目标模糊核聚类图像分割算法。该算法采用分水岭算法获得图像的过分割区域,采用多目标模糊核聚类算法对区域代表点和分水岭上的像素进行聚类。根据聚类结果将图像中的像素进行标记,得到最终的分割图像。实验结果表明,由于利用了图像区域信息,使得目标能够比较完整地从背景中分离出来。

关键词: 多目标进化算法, 核聚类, 图像分割, 分水岭