基于Gabor小波的人脸检测  

Face Detection Based on Gabor Wavelets

作　　者：聂祥飞[1] 郭军[1] 

机构地区：[1]北京邮电大学模式识别实验室

出　　处：《计算机工程》2006年第21期44-46,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60475007);教育部基金资助重点项目(02029);教育部跨世纪人才基金资助项目;重庆市教委科研项目;重庆三峡学院科研项目

摘　　要：提出了一种新的正面人脸检测算法。该方法组合了Gabor小波变换、输入图像的Gabor特征分析和Bayes分类器来进行正面人脸检测。对训练集的平均脸作Gabor小波变换得到40个投影向量;通过计算输入图像和这40个投影向量间的内积来提取图像的Gabor特征向量;训练Bayes分类器来进行正面人脸检测。实验结果表明,该算法的计算效率和检测精度均优于特征脸方法。A novel method for frontal face detection is presented. The novelty of this paper comes from the integration of the Gabor transform of mean face, face feature analysis of the input image, and the Bayes classifier for frontal face detection. 40 projection vectors are got from Gabor transform of mean face. Face feature analysis is derived from a feature vector by calculating the inner products between the input image and the 40 projection vectors. The Bayes classifier is trained to detect frontal face in an image. The experimental results show that the proposed method has lower computational complexity and higher accuracy than Eigenfaces method.

关 键 词：人脸检测 GABOR小波 BAYES分类器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]