In order to reduce the manual intervention involved in the existing segmentation methods of digital human slice images,an algorithm based on the connected component labeling and the K-means clustering is proposed. In this algorithm,first,the initial region of brain tissue is segmented via the connected component labeling and is refined via erosion. Then,a K-means clustering is adopted to extract the white matter,in which the color histogram is used to determine the clustering centers and the Euclidian distance is considered as the judging criterion. The proposed algorithm is finally applied to the segmentation of the sequential brain slice images from the first Chinese female visible human dataset. The qualitative and quantitative analyses of experimental results indicate that the proposed algorithm is of high segmentation accuracy and strong stability,and that it can be used to the automatic separation of skull from the brain tissue and to the automatic extraction of structures in deep brain.