基于离散判别函数的新的代数特征抽取方法及人脸识别  

A NOVEL ALGEBRAIC FEATURE EXTRACTION METHOD BASED ON SCATTER DISCRIMINANT FUNCTION AND FACIAL RECOGNITION

作　　者：郭跃飞[1] 邬永革[1] 黄修武[1] 姜志华[1] 杨静宇[1] 

机构地区：[1]南京理工大学计算机系,210094

出　　处：《模式识别与人工智能》1998年第3期318-322,共5页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金;国家教委博士点基金

摘　　要：本文给出了矩阵离散度的概念,并建立了图像的离散判别函数。根据这一判别函数,我们进一步给出了用于图像识别的一个新的特征抽取方法。最后用最小距离分类器对其进行识别。试验结果表明,本方法具有良好的识别性能。In this paper a scatter measure of matrix is first proposed, and a scatter discriminant function (SDF) is established. Based on the discriminant function, a novel feature extraction method for image recognition is presented. At. last, the projective feature vectors are classified with the minimum distance classifier. Experimental result shows that the present method prerforms quite well.

关 键 词：特征抽取 离散判别函数 人脸识别 图像识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]