摘要
提出一种基于半监督EM聚类的彩色图像分割方法, 算法利用了有限的人工信息, 即在图像上点击有限的几个点以标识对应区域之间的关系, 从而得到满足给定限制的精确图像分割结果。算法首先对图像进行量化处理, 而后在量化后的色彩空间中集成先验的分割信息进行色彩聚类。实验表明算法运行速度快, 分割效果好, 具有很高的应用价值。
A new image segmentation algorithm based on semi-supervised EM clustering is proposed, which makes use of limited human assistance, users indicate the relationship of some different regions in the image by mouse, to get the finally accurate segmentation result. The algorithm first has the image quantized and then clusters in the quantized color space with prior segmentation information. Experiment results show that the proposed algorithm is effective and has high value of utility.
出处
《中国体视学与图像分析》
2005年第1期50-54,共5页
Chinese Journal of Stereology and Image Analysis
关键词
图像分割
半监督聚类
高斯混合模型
image segmentation
semi-supervised clustering
mixture models
