摘要
小波变换具有良好的多尺度特征表达能力,能将图像的大部分能量集中到最低分辨率子图像,高频部分则对应于图像的边缘和轮廓,可以很好地压缩和表征人脸图像的特征。支持向量机技术针对小样本问题设计,对人脸识别这样的非线性、高维数的小样本问题有非常好的分类效果和学习推广能力,目前已经成为模式识别的首选分类器。文中使用小波变换来对人脸的高维图像矢量进行压缩,并设计了一个支持向量机分类器系统来识别人脸。试验结果验证了该系统有很高的识别率和较强的鲁棒性。
Wavelets transform has a good ability to express the multi-resolution characteristics.After the wavelets trans-form of an image,its most energy is mainly concentrated on its lowest resolution subimage.The high frequency part ex-press the boundary and profile of the image,thus we can compress and code the image using this character.Support Vector Machines(SVM)is a classifier devised for small sample problems and has an effective classify and learning a-bility.At present ,SVM is the first select classifier.This article utilizes the wavelet transform to compress the high dimen-sional face image vectors,then devises an SVM classify system to recognition the face.The try on the AT&T face database shows the effective and robust of the method.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第12期52-54,共3页
Computer Engineering and Applications
关键词
小波变换
支持向量机人脸识别
核函数
Wavelets transform,Support vector machines,Face recognition,Kernel function
