文献类型：期刊文献

英文题名：A new image color analysis method based on manifold learning

作者：Qi, Wang[1]

第一作者：王琦

通讯作者：Qi, W.

机构：[1] Department of Electrical Engineering, School of Normal, Beijing Union University, Beijing, 100011, China

第一机构：北京联合大学师范学院

通讯机构：[1]Department of Electrical Engineering, School of Normal, Beijing Union University, Beijing, 100011, China|[1141711]北京联合大学师范学院;[11417]北京联合大学;

年份：2012

卷号：43

期号：2

起止页码：187-191

外文期刊名：Journal of Theoretical and Applied Information Technology

收录：EI(收录号：20124115555612);Scopus(收录号：2-s2.0-84867179151)

语种：英文

外文关键词：Cluster analysis - Color - Dimensionality reduction - Embeddings - Gaussian distribution - Image analysis - Learning systems

摘要：In this paper, the main application image processing, manifold learning and the method of Gaussian mixture model for dimensionality reduction and cluster analysis, the image color information are all studied. First, the color data access algorithm is introduced, secondly, the manifold learning in the local linear embedding (LLE) algorithm is used in color analysis; then the results of an evaluation criteria, LLE parameters in the criteria automatic selection algorithm are presented; meanwhile, the result of the operation of the different color space LLE are also tested and analyzed. Finally, the application of a Greedy EM-based Gaussian mixture model for improving the operation of the HSI space under LLE results of experiments is analyzed. It indicates that the algorithm can automatically determine the number of clusters, and achieve a better clustering result. ? 2005 - 2012 JATIT & LLS. All rights reserved.