摘要
基于k-Means等聚类算法的图像分割对聚类中心的初始选择敏感,可靠性差。为避免初始聚类中心选择的影响,将相似性传播聚类用于灰度图像分割。另外,为降低该聚类算法输入相似度矩阵的计算时间复杂度,提出用待分割图像中出现过的灰度值代替像素点作为数据点进行聚类。实验结果表明,与基于k-Means聚类的分割算法相比,该算法不需要预设聚类中心,可靠性更高。
The image segmentation methods based on the k-Means clustering are highly sensitive to the cluster centers' initialization. To avoid the effects of the cluster centers' initialization, the affinity propagation clustering was used to sgement grey images. To reduce the complexity in calculating the input similarity matrix of the affinity propagation clustering, it was proposed that all grey levels were selected to he used as data points instead of all pixels in the image. The results show that the method proposed need not preelect cluster centers and is more reliable than the method based on the k-Means culstering.
出处
《海军工程大学学报》
CAS
北大核心
2009年第3期33-37,共5页
Journal of Naval University of Engineering
基金
国家自然科学基金资助项目(60804034)
山东科技大学科学研究"春蕾计划"资助项目(2008BWZ054)
关键词
相似性传播聚类
图像分割
聚类中心
灰度值
时间复杂度
affinity propagation clustering
image segmentation
cluster center
grey level
time complexity