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.
出处
《自动化学报》
EI
CSCD
北大核心
1992年第2期233-238,共6页
Acta Automatica Sinica
关键词
图象识别
代数特征抽取
奇异值特征
Image recognition
algebraic feature extraction
singular value feature
human facial recognition
discriminant vector