| 基于拉普拉斯图的谱特征的图像聚类研究 |
| |
| 引用本文： | 孔敏,汤进,罗斌.基于拉普拉斯图的谱特征的图像聚类研究[J].中国科学技术大学学报,2007,37(9):1125-1129. |
| |
| 作者姓名： | 孔敏 汤进 罗斌 |
| |
| 作者单位： | 1. 皖西学院机械与电子工程系,安徽六安,237012;安徽大学计算机科学与技术学院,安徽合肥,230039
2. 安徽大学计算机科学与技术学院,安徽合肥,230039 |
| |
| 基金项目： | 国家自然科学基金;安徽省教育厅自然科学基金 |
| |
| 摘 要： | 提出一种用拉普拉斯图的谱系数夹角谱特征来描述图像几何结构的方法,同时研究了基于图的谱聚类系统.首先将序列图像以角点的形式构成拉普拉斯矩阵;然后分解该矩阵,结合特征值和其特征向量计算图中各点的谱系数夹角谱特征;再以局部保持投影方法将这些向量内嵌到模式空间,并在其特征空间用模糊c-均值算法进行聚类分析.结果表明,以拉普拉斯图的谱系数夹角谱特征解决了图中各点在向量空间的分布及其对应关系,在模式空间进行的聚类分析是有效的.
|
| 关 键 词： | 谱聚类 谱系数夹角 模糊c-均值 局部保持投影 聚类有效性 |
| 文章编号： | 0253-2778（2007）09-1125-05 |
| 修稿时间： | 2007-01-26 |
Image clustering based on spectral features of Laplacian graph |
| |
| KONG Min,TANG Jin,LUO Bin.Image clustering based on spectral features of Laplacian graph[J].Journal of University of Science and Technology of China,2007,37(9):1125-1129. |
| |
| Authors: | KONG Min TANG Jin LUO Bin |
| |
| Institution: | 1. Machine and Electron Engineering Department, West Anhui University, Luan 237012, China ; 2. School of Computer Science and Technology, Anhui University, Hefei 230039, China |
| |
| Abstract: | A geometric structure representation of images based on the angle between spectral coefficient vectors of Laplacian graph was proposed.Graph clustering of spectral features was also investigated.Laplacian matrix was constructed by the corner points of oriented graph.Then the angles between spectral coefficients vectors were computed based on eigenvalues and eigenvectors after the matrix was decomposed.Lastly,fuzzy c-mean clustering(FCM) was performed on the pattern space embedded by locality preserving projections(LPP).Experimental results show that the spectral features of the angles between spectral coefficient vectors can describe the distribution and relationship of all graph nodes and the clustering in this feature pattern space is valid and feasible. |
| |
| Keywords: | spectral clustering spectral coefficient angle fuzzy c-mean clustering locality preserving projections clustering validity |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
