通过对相似性度量函数的研究,得到了参数σ的取值与聚类类数间的关系,并据此提出了一种自动确定参数σ的方法,设计了一种基于相似度因子的无监督聚类算法(unsupervised clustering algorithm based on similarity parameter,UCASP),在图像分割的实际应用中实现了无监督分割。
This paper found the affiliation between the parameter and the number of cluster on the basis of the research of similarity function. Meanwhile, achieved an approach to assigning parameter σ automatically. Finally, developed an unsupervised clustering algorithm, which achieved an really unsupervised segmentation on real-world data.