用于图象识别的图象代数特征抽取  

ALGEBRAIC FEATURE EXTRACTION OF IMAGES FOR RECOGNITION

作　　者：洪子泉[1] 杨静宇[1] 

机构地区：[1]华东工学院计算机系,南京210014

出　　处：《自动化学报》1992年第2期233-238,共6页Acta Automatica Sinica

摘　　要：本文证明了图象的奇异值特征具有一系列代数和几何上的不变性以及对噪音的不敏感性,它是识别图象的有效特征。本文将奇异值特征用于人象识别问题。根据图象奇异值特征向量样本建立了Sammon最佳鉴别平面上的正态模式Bayes分类模型。实验结果表明,奇异值特征向量具有良好的鉴别分离能力。In this paper, it is proved that the Singular Value (SV) feature vector has some important properties such as the invariance to the algebraic and geometric transformations, and the insensitiveness to noise, Therefore, th SV feature is very useful for describing and recognizing images. As an example, the SV feature vector is used to a problem of recognizing human facial images. The normal pattern Bayes classification model based on the Sammon optimal discriminant plane is constructed. The experimental results show that the SV feature vector has strong ability for the separating classes.

关 键 词：图象识别 代数特征抽取 奇异值特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]