摘要
从最优化的角度出发,提出了一种基于分块小波变换和二维主成分分析法(2DPCA)的人脸特征提取与识别算法。该方法首先对人脸图像进行分块小波变换,并对各分块的高、低频分量进行组合处理,然后对小波系数特征应用2DPCA方法进行变换并将分块特征进行融合得到人脸鉴别特征,最后在ORL人脸库上应用支持向量机(SVM)对该特征进行分类识别。试验结果表明,该算法能有效地提高人脸识别性能,具有较短的识别时间和较高的识别准确率,优于传统的人脸识别方法。
How to extract face recognition information from an image was investigated in this paper. A feature extraction and recognition algorithm of intersected human face based on wavelet transform and 2DPCA was proposed, by which recognition features of an image could be easily extracted for a discriminant method for face recognition. Firstly, the intersected huaman face was transformed with wavelet and different coefficients were extracted. Wavelet coefficient features were gotten by combing low frequency coefficients with high frequency coefficients of each block. Then, the 2DPCA method was used to extract features from wavelet coeficinent features and fused into ultimate discriminant features. Finally, the features were classified and recognized by SVM. The efficiency of proposed algorithm was validated.Experiment results demonstrate that the proposed method is not only good at recognition speed, but also achives a higher accuracy than classical methods.
出处
《红外与激光工程》
EI
CSCD
北大核心
2012年第11期3118-3122,共5页
Infrared and Laser Engineering
基金
山东省自然科学基金(ZR2010FM023)
