| 用于人脸识别的张量图优化的Fisher判别分析 |
| |
| 引用本文： | 单桂军. 用于人脸识别的张量图优化的Fisher判别分析[J]. 科学技术与工程, 2013, 13(15): 4212-4216 |
| |
| 作者姓名： | 单桂军 |
| |
| 作者单位： | 江苏科技大学 |
| |
| 基金项目： | 江苏省高校实验室研究会研究课题（JS2012-2），江苏省现代教育技术研究2010年度课题（16866） ，镇江市科技支撑计划项目（GY2012041） |
| |
| 摘 要： | 最近提出的图优化的Fisher判别分析(Graph-based Fisher Analysis,简称GbFA)具有很强的判别性能,并成功地应用于人脸识别。但GbFA需要将二维的人脸图像矩阵转化为向量,因此容易丢失像点的空间关系。为此,提出用于人脸识别的张量图优化线性判别分析(Tensor Graph-based Fisher Analysis,简称TGbFA)。该算法把二维的人脸图像矩阵看作二维张量数据,并通过GbFA方法迭代求得两个投影矩阵。在Yale和YaleB的人脸库的实验表明,TGbFA算法继承了GbFA的特性,与现有的张量线性判别分析算法相比,TGbFA具有较好的判别性能。
|
| 关 键 词： | 人脸识别 降维 图优化的Fisher判别分析 张量图像 |
| 收稿时间： | 2013-01-14 |
| 修稿时间： | 2013-02-28 |
Tensor Graph-optimized Fisher Discriminant Analysis for Face Recognition |
| |
| shanguijun. Tensor Graph-optimized Fisher Discriminant Analysis for Face Recognition[J]. Science Technology and Engineering, 2013, 13(15): 4212-4216 |
| |
| Authors: | shanguijun |
| |
| Affiliation: | SHAN Gui-jun(School of Electronic Information,ZhenJiang University of Science and Technology,Zhenjiang 212003,P.R.China;Department of Electronic and Information,ZhenJiang College,Zhenjiang 212003,P.R.China) |
| |
| Abstract: | Latest proposed Graph-based Fisher Discriminant Analysis (GbFA) for dimensionality reduction has powerful discriminant ability and been applied successfully face recognition. However, GbFA transforms two-dimensional face image matrixes into vectors, which lose spatial relations of image pixels. Tensor Graph-based Fisher Analysis (TGbFA) is proposed, which regards face image matrixes as two-order tensor data and gets two projection matrixes through the iteration way. Experiments on Yale and YaleB face datasets show that TGbFA inherits the propriety of GbFA and the algorithm has more discriminant than other existing tensor linear discriminant analysis. |
| |
| Keywords: | Face Recognition Dimensionality Reduction Graph-based Fisher Discriminant Analysis Tensor Image |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《科学技术与工程》浏览原始摘要信息 |
|
点击此处可从《科学技术与工程》下载全文 |
|
