计算机工程与应用
計算機工程與應用
계산궤공정여응용
COMPUTER ENGINEERING AND APPLICATIONS
2014年
8期
158-160,187
,共4页
人脸识别%矩阵主成分分析%小波包
人臉識彆%矩陣主成分分析%小波包
인검식별%구진주성분분석%소파포
face recognition%2-Dimensional Principal Component Analysis(2DPCA)%wavelet packets
提出一种改进的小波包融合+2DPCA方法,先对图像进行二层小波包分解,再选取最利于判别分类的4幅高频子图进行融合,将融合子图与低频子图分别进行2DPCA降维和特征提取,最后进行决策级融合,得到识别结果。在Yale和JAFFE标准人脸库上的实验结果表明,该改进方法能有效提高识别率。
提齣一種改進的小波包融閤+2DPCA方法,先對圖像進行二層小波包分解,再選取最利于判彆分類的4幅高頻子圖進行融閤,將融閤子圖與低頻子圖分彆進行2DPCA降維和特徵提取,最後進行決策級融閤,得到識彆結果。在Yale和JAFFE標準人臉庫上的實驗結果錶明,該改進方法能有效提高識彆率。
제출일충개진적소파포융합+2DPCA방법,선대도상진행이층소파포분해,재선취최리우판별분류적4폭고빈자도진행융합,장융합자도여저빈자도분별진행2DPCA강유화특정제취,최후진행결책급융합,득도식별결과。재Yale화JAFFE표준인검고상적실험결과표명,해개진방법능유효제고식별솔。
An improved method based on wavelet packets fusion and 2DPCA is proposed. Firstly, the original face image is decomposed by wavelet packets at two levels, and four most conducive high-frequency sub-images are selected and fused to improve the performance of classification. Then, 2DPCA is carried out in high-frequency and low-frequency sub-graphs separately. Finally, the decision level fusion is used to get the recognition result. Experimental results show that the proposed method is effective to face recognition with Yale and JAFFE databases.