Please use this identifier to cite or link to this item:
標題: 人臉識別特徵萃取演算法之研究
A Comparative Study on Feature Extraction for Face Recognition
作者: 張嘉峰
Chang, Chia-Feng
關鍵字: Face recognition
Feature extraction
Principle component analysis
Wavelet transform
Feature selection
出版社: 資訊科學與工程學系所
引用: [1]. M. Turk, A. Pentland, “Eigenfaces for recognition.”, J. Cognitive Neuroscience. 3 (1) (1991) 71-86. [2]. P.N. Belhumeur, J.P. Hespanha and D.J. Kriegman, “Eigenfaces vs. Fisherface: Recognition using class specific linear projection.”, IEEE Trans. Pattern Anal. Machine Intell. 19 (7), pp. 711-720,1997. [3]. Keinosuke Fukunaga , “Introduction to statistical pattern recognition”,1990. [4]. J. Y, D. Zhang, A.F. Frangi, J.Y. Yang,”Two dimensional PCA: a new approach to appearance based face representation and recognition.”, IEEE Trans. Pattern Anal. Mach. Intell. 26 (1) ,2004. [5]. D. Zhang, Z.H. Zhou,”(2D)2PCA: two-directional two-dimensional PCA for efficient face representation and recognition.”, Neurocomputing 69 (2005) 224-231. [6]. Young-Gil Kim, Young-Jun Song, Un-Dong Chang, Dong-Woo Kim,” Face recognition using a fusion method based on bidirectional 2DPCA.”, Applied Mathematics and Computation 205 (2008) 601-607. [7]. Te-Ming Tu, Chin-Hsing Chen, Jiunn-Lin Wu, and Chein-I Chang,” A Fast Two-Stage Classification Method for High-Dimensional Remote Sensing Data.” IEEE Trans. Geosciemce and Remote Sensing, Vol. 36, No. 1, Jan. 1998 [8]. Ping Dong, Jovan G. Brankov, Nikolas P. Galatsanos , Yongyi Yang, and Franck Davoine,” Digital Watermarking Robust to Geometric Distortions.”, IEEE Trans. on Image Processing, 14 (12), Dec. 2005 [9]. R. C. Gonzalez and R. E. Woods, “Digital Image Processing” 2002 edition. [10]. The ORL Face Database, [11]. The Yale Face Database, [12]. I. Daubechies,” The wavelet transform, time-frequency localization and signal analysis.”, IEEE Trans. Information Theory, Vol. 36, No. 5, 961-1005, 1990
摘要: 人臉識別是一種利用每張人臉圖片的資訊,經過特徵萃取的動作,利用這些特徵來達成識別的工作。臉部辨識主要有兩個挑戰第一人臉影像會有不同的表情和姿勢當我們變換表情和姿勢時會使的臉部辨識的工作更加困難。第二影像中有不均勻的光影照射也會影像辨識的結果。對於第一個挑戰,有人提出主要成份分析方法去解決。主要成份分析(PCA)人臉識別方法,在光影的影響下會造成許多的誤判,並且在我們的訓練資料集中的資料有所改變的時候,就必須重新計算投影軸,進而產生了要一直重複訓練的缺點。 因此本論文提出一個有效的人臉識別演算法,其結合小波轉換和雙向二維主成份分析的人臉辨識方法,能有效去除影像中像雜訊的部份。在這邊我們使用小波轉換去除影像中像雜訊的部份。接著為了加速人臉辨識的時間和其辨識成功率我們把處理完的影像使用雙向二維主成份分析的方法取出包括雙向影像結構資訊的特徵,然後使用特徵選取的方法取出較具有代表性的特徵拿來分類,最後使用k個最近鄰居的分類器作分類。實驗結果顯示所提演算法能解決我們所描述的問題而有一個不錯的識別正確率。
Face recognition is an important but complex problem which has been widely used in many fields, such as surveillance, security and telecommunications. The complexities of face recognition mainly lie in the changing appearance of human face, such as variations in illumination, posture and expression. To overcome the recognition problem of faces with varying expression and posture, the PCA-base face recognition had presented to solve it. Generally, the PCA-based method is utilized in the feature extraction process to reduce the dimensionality of the original face images. However, it has the following problems: (1) the extracted features are global features for all face classes and thus may not be optimal for discriminating one face class from the others; (2) the computation of principal components are data dependent; (3) it does not work well for recognizing human faces under different illumination. A new feature extraction method for face classification is presented in this study. Wavelet Transform is used to remove noise-like pixel from the face images; it eliminates the effect of noise. In order to speed up the classification, (2D)2PCA method is performed for feature reduction. The extracted features using the proposed method are robust to varying illumination, different posture and expression changes. Then picking important features by feature selection method; it efficient reduces the number of features and improve the performance. Finally, nearest-neighbors based classifier is used for face recognition. Expected experimental result show that theproposed method achieves satisfactory classification result.
其他識別: U0005-1607201015160700
Appears in Collections:資訊科學與工程學系所



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.