中文摘要：

提出了一种基于分水岭变换和核聚类算法的图像分割算法．通过分水岭变换把图像分割成多个小区域，为实现过分割小区域的合并，利用Mercer核把各小区域的灰度平均值映射到高维特征空间，使原来没有显现的特征突现出来，在特征空间进行更准确的聚类，为下一步图像分析提供较为准确的分割区域．实验结果证明了该算法的可行性和有效性．

英文摘要：

A method of image segmentation algorithm based on watershed translation and kernel clustering is presented. The image is segmented into a large number of small regions by the watershed translation. Mercer kernel which will map the mean grey-level values of these small regions to a high dimensional feature space is employed to merge these over-segmented regions. This mapping makes the unconspicuous feature obvious, the clustering in the feature space more precise and provides the segmented regions for further image analysis precisely. The feasibility and effectiveness of the algorithm is validated by the experimental results.