Abstract: In view of face recognition with only one sample problem, this paper presents a method of face recognition based on virtual image. In order to enhance the classification information of single training sample, it first adds some virtual images generating from the given training image. By using Discrete Wavelet Transform(DWT), the low-frequency band is used as feature for recognition. Two-dimensional Principal Component Analysis(2DPCA) is performed on the low-frequency faces. Experimental results show that this method can filter the high frequency of informations which are caused by a expression change and little cover, and promote the recognition rate.

Key words: virtual image, Discrete Wavelet Transform(DWT), single sample, face recognition

摘要： 针对单训练样本情况下的人脸识别问题，提出一种基于虚拟图像的人脸识别方法。为给定的训练图像增加虚拟图像，以增强单训练样本的分类信息，对其进行离散小波变换，并将变换的低频子带图像作为人脸识别特征，利用二维主成分分析法分析“低频脸”。实验结果表明，该方法能过滤因表情变化和少量遮掩而带来的高频信息，提高识别率。

关键词: 虚拟图像, 离散小波变换, 单样本, 人脸识别

CLC Number: 

• TP391.9