圖形識別和電腦視覺在人臉辨識問題上有日益增多的相關研究。現今的系統也已發展出相當精準的辨識率，但是仍然無法克服影像上較大的各種變化，例如:觀看的方向或是姿勢、臉部表情、光線照明狀況、人類的變老，以及裝扮(臉部頭髮、眼鏡、化妝)。

　　本論文提出2DPCA+GF的人臉辨識方法。此方法主要是依據二維主軸分析(2DPCA)及Gabor特徵。在方法裡，每一原始影像先跟40個不同方向及尺度的Gabor濾波器做捲積得到其Gabor影像，然後使用2DPCA對Gabor 影像作二維主軸分析。二維主軸分析沒有將Gabor 影像轉換成向量即直接對其求得共變異(covariance)矩陣，因而使得計算效率大為提高。本論文提出的2DPCA+GF方法結合了2DPCA和Gabor濾波器的優點。本論文並探討一個不同但相似於2DPCA+GF的方法，稱為2DPCA+MGF，2DPCA+MGF方法以原始空間域影像取代部份的Gabor影像成為一種混和方法。

　　本論文使用ORL資料庫對PCA、2DPCA、2DPCA+GF和2DPCA+MGF四種方法進行比較實驗，使用的是1-norm和2-norm兩種最短距離分類器。前面兩種方法(PCA,2DPCA)早為前人研究過，後兩種方法(2DPCA+GF,2DPCA+MGF)則是本篇論文的新貢獻。實驗結果顯示：2DPCA+MGF方法使用1-norm距離量測得到的辨識率比使用2-norm距離量測來得好。使用25個2DCPA主軸及1-norm最短距離分類法，2DPCA+MGF可以達到辨識率98.5%，2DPCA+GF是93%，2DPCA是90.5%，PCA就滑落到76.5%而已。實驗結果進一步證實Gabor表示比空間域表示更能掌握分辨資訊，而2DPCA比PCA在辨識率及實現複雜度上均較優勢。

　Pattern recognition and computer vision have witnessed the growing interests in face recognition problems. Current systems have advanced to be fairly accurate in recognition. But they still unable overcome the large variations, such as viewing directions or poses, facial expression, illumination conditions, aging, and disguises (facial hair, glasses or cosmetics).

　This thesis presents a new face recognition method based on Two-Dimensional Principal Component Analysis (2DPCA) and Gabor filters. In the method, an original image is convolved with 40 Gabor filters corresponding to various orientations and scales to give its Gabor representation. Then, the Gabor representation is analyzed by the 2DPCA in which the eigenvectors are computed using the Gabor image covariance matrix without matrix-to vector conversion. The proposed 2DPCA+GF method combines the advantages from 2DPCA and Gabor filters. A different version of the 2DPCA+GF, called 2DPCA+MGF, is also studied. In the 2DPCA+MGF, some of Gabor images are replaced by the original spatial-domain image to give a mixture representation.

　Experiments based on the ORL database were then performed to compare the recognition rate between the PCA, the 2DPCA, the 2DPCA+GF and the 2DPCA+MGF methods using the 1-norm and 2-norm minimum distance classifiers. The former two methods (PCA and 2DPCA) were studied by others before. The study on the latter two methods (2DPCA+GF and 2DPCA+MGF) is our new contribution. We find that the recognition rate using 1-norm distance measure is better than the 2-norm measure in the 2DPCA+MGF method. It achieves 98.5% recognition rate by using 25 principal components of 2DPCA using the 1-norm distance classifier. Under the same condition, the 2DPCA+GF achieves 93% recognition rate, the 2DPCA achieves 90.5% recognition rate, the PCA achieves 76.5% recognition rate. This study further confirm that the Gabor representation carries more discriminating information than its counterpart, the spatial-domain representation and the 2DPCA has advantage over the PCA both in recognition rate and implementation complexity.

[1] D. Burr, M. Morrone, and D. Spinelli, “Evidence for Edge and Bar Detectors in Human Vision,” Vision Res, vol. 29, no. 4, pp. 419-431, 1989.
[2] P. N. Belhumeur, J. P. Hespanha, and D. J. Kriegman, “Eigenfaces vs. Fisherfaces: Recognition Using Class Specific Linear Projection,” IEEE
Trans. on Pattern Analysis and Machine Intelligence, vol. 19, no. 7, pp. 711-720, Jul. 1997.
[3] Ki-chung Chung, Seok Cheol Kee, Sang Ryong Kim, “Face Recognition Using Principal Component Analysis of Gabor Filter Responses,” Proceedings.
International Workshop, pp. 53– 57, Sept. 1999.
[4] J.G. Daugman, “Two-Dimensional Spectral Analysis of Cortical Receptive Field Profiles,” Vision Research, vol. 20, pp. 847–856, 1980.
[5] J. Daugman, “Uncertainty Relation for Resolution in Space, Spatial Frequency, and Orientation Optimized Two-Dimensional Visual Cortical Filters,” Journal Opt. Soc. Am, pp. 1160-1168, 1985.
[6] I. Daubechies,“The Wavelet Transform, Time-Frequency Localization and Signal Analysis,” IEEE Trans. Inf. Theory 36, pp. 961—1005, 1990.
[7] G. Donato, M. S. Bartlett, J.C. Hager, P. Ekman, and T. J. Sejnowski,“Classifying Facial Actions,” IEEE Trains. Pattern Anal. Machine Intell,
[8] D. Gabor, “Theory of Communication,” J. Inst. Electr. Eng., Vol. 93, 1946.
[9] A.L. Graps, “An Introduction to Wavelets,” IEEE Computational Sciences and Engineering, Volume 2, Number 2, Summer, pp. 50-61. 1995.
[10] C. Garcia, G. Zikos, G. Tziritas, “Wavelet Packet Analysis for Face Recognition,“ Image and Vision Computing, Vol. 18, pp. 289-297, 2000.
[11] Rafael C. Gonzalez and Richard E. Woods, “Digital Image Processing,”Second Edition, Prentice Hall, NJ, 2002.
[12] Jia-Hui Guo, “Wavelet Bookmark”, Department of Mathematics National Central University <http://www.math.ncu.edu.tw/~guo/ wavelet_bookmark/index.html>.
[13] A. Hyvrinen, “Fast and Robust Fixed-Point Algorithms for Independent Component Analysis,“ Neural Computing Surveys, vol. 2, pp. 94-128, 1999.
[14] X. He, P. Niyogi, “Locality Preserving Projections,” Proc. Conf. Advances in Neural Information Processing Systems, 2003.
[15] X. He, S. Yan, Y. Hu, P. Niyogi, H. Zhang, “Face Recognition Using Laplacianfaces,” IEEE Trans. on Pattern Analysis and Machine
Intelligence, Volume 27, Issue 3, pp. 328-340, March 2005.
[16] I.T. Jollife, “Principal Component Analysis,” Springer-Verlag, New York 1986.
[17] J. Jones and L. Palmer, “An Evaluation of The Two-Dimensional Gabor Filter Model of Simple Receptive Fields in Cat Striate Cortex,” J.
Neurophysiol, pp. 1233-1258, 1987.
[18] M. Lades, J.C. Vorbruggen, J. Buhmann, J. Lange, C. von der Malsburg, Wurtz R.P., and W. Konen, “Distortion Invariant Object Recognition in
The Dynamic Link Architecture,” IEEE Trans. on Computers, vol. 42, pp. 300–311, 1993.
[19] C. Liu, H. Wechsler, “Gabor Feature Based Classification Using the Enhanced Fisher Linear Discriminant Model for Face Recognition,” IEEE
Trans. Image Processing, vol. 11, no. 4, pp. 467-476, 2002.
[20] P.S. Penev, L. Sirovich, “The Global Dimendionality of Face Space,” Proc. Fourth IEEE Int’l Conf. Automatic Face and Gesture Recognition, pp. 264-270, 2000.
[21] K. Sandberg, “The Haar Wavelet Transform,” <http://amath.colorado.edu/courses/4720/2000Spr/Labs/Haar/haar.html>. Online, April 1, 2000.
[22] M. Turk, A. Pentland, “Eigenfaces for Recognition,” J. Cognitive Neuroscience, vol. 3, no. 1, pp. 71-86, 1991.
[23] L. Wiskott, J. M. Fellous, N. Kruger, and C. von der Malsburg, “Face Recognition by Elastic Bunch Graph Matching,” IEEE Trans. on Pattern
Analysis and Machine Intelligence, vol. 19, pp. 775-779, 1997.
[24] D. Xu, S. Yan, L. Zhang, M. Li, W. Ma, Z. Liu, H. Zhang, “Parallel Image Matrix Compression for Face Recognition,” Multimedia Modelling Conference, 2004. MMM. Proceedings of the 11th International 12-14 Jan. pp. 232-238, 2005.
[25] M. H. Yang, N. Ahuja, and D. Kriegman, “Face Recognition Using Kernel Eigenfaces,” in proc. IEEE Int. Conf. Image Processing, Sept. 2000.
[26] J. Yang, D. Zhang, et al. “Two-dimensional PCA: a New Approach to Appearance-Based Face Representation and Recognition,” IEEE Trans. on
Pattern Analysis and Machine Intelligence, vol. 26, pp. 131-137, 2004.
[27] B. Zhang, W. Gao, S. Shan, W. Wang “Constraint Shape Model Based On Edge Constraint and Gabor Wavelet Search,” AVBPA, pp. 52-63, 2003.