摘要
提出一种基于分块离散余弦变换(DCT)与奇异值分解阈值压缩(TCSVD)的人脸特征提取与识别算法。该算法对人脸图像进行分块DCT变换,根据图像块位置和能量分布选择不同的DCT高低频分量构建特征矩阵,通过对每个图像块的特征矩阵进行SVD阈值压缩和特征组合来构建人脸鉴别特征,并利用分类器进行特征分类与识别。AR人脸库上的实验结果验证了该算法的有效性。
A human face feature extraction and recognition algorithm based on Divided Discrete Cosine Transform（DDCT） and Singtflar Value Decomposition Threshold Compression（TCSVD） is proposed. With this method, the human faces are divided into different sub-images and transformed with DDCT, and different high and low frequency coefficients are chosen to construct DCT feature matrix. TCSVD algorithm is used to extract the ultimate discriminative features, And the features are classified with the feature classifier. Experimental results on AR face database validate the efficiency of the method.
出处
《计算机工程》
CAS
CSCD
北大核心
2008年第9期4-6,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60472060,60472061,60473039)
关键词
人脸分割
分块离散余弦变换
奇异值分解阈值压缩
特征提取
human face division
Divided Discrete Cosine Transform（DDCT）
Singular Value Decomposition Threshold Compression（TCSVD）
feature extraction
