Adaptive K-means clustering for color image segmentation

作者：Yong Zhou; Shi Haibin^*

来源：Advances in Information Sciences and Service Sciences, 2011, 3(10): 216-223.

DOI：10.4156/AISS.vol3.issue10.27

• Cluster validity indices
• Clustering number
• Color distribution
• Color image segmentation
• Color images
• Color quantization
• Human intervention
• Initial cluster centers
• k-Means algorithm
• K-means clustering
• K-Means clustering algorithm
• Octree
• Optimal clustering

According to the indeterminate clustering number k and initial cluster center in K-means clustering algorithm while used in color image segmentation, a new segmentation approach by adaptive K-means clustering algorithm was proposed. The major improvements include: Firstly, adopt Octree color quantization algorithm to quantize the color image, shown with representative characteristics; then, apply a mean-equivalent method to determine initial cluster center, based on the color distribution of image; finally, cluster with K-means algorithm, and define the optimal clustering number adaptively through a new cluster validity index designed in this paper. Experimental results show that the proposed algorithm not only gives more accurate clustering number than the other algorithm, but also be effective, which greatly reduces the human intervention and has high practical value.

• 出版日期2011
• 单位中国矿业大学（北京）