摘要
研究人脸识别优化问题,人脸图像受光照、人脸表情和位置变化等因素影响,由于图像具有复杂的多尺度特征,传统人脸识别算法只能提提取局部或全局特征,不能准确描述人脸图像,导致人脸识别率低。为了提高人脸识别率,提出一种小波分解和LBP算子相结合的人脸识别算法(WTLBP)。WTLBP首先利用小波变换对人脸图像进行分解,将人脸图像分解成大尺度和小尺度图像,然后采用LBP算子提取人脸图像的多尺度特征,最后采用概率统计法对人脸进行匹配识别。对ORL人脸库进行仿真,结果表明,WTLBP能够提取到人脸图像更加丰富的局部和全局信息,对光照、人脸表情和位置变化具有较高的鲁棒性,提高了人脸识别率。
The paper researched the problem of face recognition. Face image is affected by the factors of beam, expression and location, etc, and it has characters of complex muhi-scale. Traditional face recognition algorithm can only exact the face' s local or global characteristics, but it cannot describe the face image accurately, which leads to the low accuracy of recognition. In order to improve the accuracy of recognition, we proposed an algorithm in which wavelet decomposition was combined with LBP operator (WTLBP). Firstly, it decomposed face image with wavelet transform and divided the face image into large scale and small scale images. Then, LBP operators were adopted to extract the multi-scale characters of face image. At last, it recognized the face by the means of probability statistics. The simulation experiments were carried out based on ORL face library, and the results show that WTLBP can extract the more local and global information of face image, and it has the strong robustness against the change of illumination, expression and location. It has improved the face recognition rate markedly.
出处
《计算机仿真》
CSCD
北大核心
2012年第1期209-213,共5页
Computer Simulation
基金
辽宁省重点实验室资助项目(2008s115)
关键词
人脸识别
小波变换
局部二值模式
多尺度
直方图
Face recognition
Wavelet decomposition
Local binary mode
Multi-scale
Histogram