Adaptive Clustering Segmentation Based on Pixel Color
Jizhao Hua, Jianguo Wang · 2009
On the basis of the two limitations of cluster algorithm, an adaptive algorithm to accurately segment the color images is proposed. At first, by analysis of color's distribution in an image, the center of pixels' Hue can be found. Taking these hues as main hues, the initial cluster centers are made of the most far adjacent pixels. Then pixels are clustered by the algorithm using color and spatial information. During the alternation, usage of color and geometrical measure assure the algorithm fast and accurate. At last, the pixels are quantized according to the clustering results. In the algorithm, the value of k is established by the algorithm itself, and it changes in the alternation automatically. The clustering results are consisted of the object areas. Experiments show the method can effectively achieve the desired color image segmentation.