多尺度对称变换及其应用于定位人脸特征点  

Multi-Scale Symmetry Transform with Application to Location of Feature Points on Human Face Image

作　　者：彭进业[1] 俞卞章[1] 王大凯[2] 李楠[2] 

机构地区：[1]西北工业大学电子工程系,陕西西安710068 [2]西北大学电子科学系,陕西西安710069

出　　处：《电子学报》2002年第3期363-366,共4页Acta Electronica Sinica

基　　金：陕西省科技研究发展计划项目 (No .8k0 7 G4);西安市科技攻关计划项目 (No .GG990 7)

摘　　要：提出了一种在图像的反对称双正交小波分解数据域中 ,实现多尺度对称变换的方法 ,并将它应用于脸部图像中主要特征点的定位 .这种方法与直接在原始图像数据域中的同类方法相比 ,能大幅度减少计算量、加快定位速度 ,并明显提高定位准确度 .除此之外 ,由于它是在小波分解数据域上实现的 。A method to implement multi scale symmetry transform in the wavelet coefficients domain decomposed by using anti symmetrical biorthogonal wavelets (ASBW) is presented with application to location of feature points on human face image.Compared to the original symmetry transform,which is performed on a single resolution level,our method can yield results with higher location accuracy without increment of computational burden.In addition,if an image is compressed based on ASBW decomposition,then our method provides a way to extract image features in compressed data domain without need of completely reconstructing the image.This is one of the most important technical consideration in the content based indexing (CBI) problem.

关 键 词：多尺度对称变换 正交小波 人脸特征 图像识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]