摘要
在利用传统局部方向模式(LDP)及其改进方法进行人脸识别时,存在两个主要问题,一是局限于3×3邻域来提取特征,二是在人脸识别时将图像不同分块同等对待。针对上述问题,提出了改进的LDP。一方面提出了多分辨率分析方法,将LDP由3×3窗口模式扩展至多分辨率窗口模式;另一方面,通过采用分块加权法来增强不同分块对人脸识别的不同贡献程度。采用Yale和Orl人脸图像库进行实验,结果表明:基于改进的方法明显提高了人脸识别的效果。
While using traditional local directional pattern （LDP） and its improved method to carry out face recognition,there exist two problems, one is feature extraction is confined to 3 × 3 neighborhood;another is regions of face image are treated equally for face recognition. An improved LDP is proposed to handle the problems. On one hand, multi-resolution analysis method is proposed. LDP is extended from 3 × 3 window pattern to multi-resolution window pattern ; on the other hand, different degree of contribution blocking to face recognition is enhanced by using block weighted method. Two face image databases, Yale and Orl, are used for experiment, and experimental results show that the introduced scheme can improve effect of face recognition.
出处
《传感器与微系统》
CSCD
2016年第6期52-55,共4页
Transducer and Microsystem Technologies
基金
河南省国际合作项目(134300510057)
河南省基础与前沿技术研究项目(132300410462)
关键词
人脸识别
局部方向模式
加权
多分辨率
face recognition
local directional pattern （LDP）
weighted
multi-resolution
