本篇論文提出一種利用人臉區塊特徵來提升人臉辨識準確率的方法。在光線變化強烈的環境中，如果利用整張人臉影像作辨識，由於膚色區域極容易受到光線變化影響導致辨識率下降，故本篇論文提出以人臉五官區塊取代整張人臉的的辨識方法。
本系統的演算法首先是利用Active Appearance Model(AAM)偵測出所需要的五官影像，分別為左眉毛、右眉毛、左眼、右眼、鼻子及嘴巴這六個components。取出這六個區塊之後，使用Principal Component Analysis(PCA) 個別抽取五官的特徵向量同時降低資料維度，將這些特徵向量利用K-means方法分群，再使用Support Vector Machine(SVM)將這K個類別的資訊訓練出一個分類模組。最後辨識時，將各個component的辨識結果利用投票法整合。
實驗是由自行拍攝的影像資料庫做測試，此資料庫為有光線變化的資料庫，包含日光燈和桌燈的光源變化，以及這兩種光源從不同角度照射的光方向性變化。實驗結果顯示，當訓練影像包含所有光線變化的影像時，EigenFace的辨識率為94%，本系統的辨識率達96%。但當訓練影像為光線變化較不明顯而測試影像為光線變化較強烈時，EigenFace的辨識率為31%，相對於本系統的辨識率為63%。由實驗結果得知，本論文提出以五官影像為特徵來提升人臉辨識率的方法較不需要使用特殊光線變化情況下的人臉影像做訓練，對於訓練資料集具有較大的容忍力。相較於EigenFace的辨識率，也有顯著的提升。

In this thesis, we present a component-based face recognition method using the facial block feature to increase the recognition rate. In the complex lighting environment, if the system takes the whole facial images for recognition, the skin area will be extremely easy to be influenced by the lighting changes so as to decrease the recognition rate. To remedy this problem, we propose a method using the facial components images instead of the whole facial images for recognition.
Firstly, the Active Appearance Model (AAM) is adopted to detect six facial components images, which are left eyebrow, right eyebrow, left eye, right eye, nose and mouth, respectively. Then, Principal Component Analysis (PCA) is utilized to calculate the desired feature vectors and decrease the dimension of the original feature vectors. After that, the K-means algorithm is employed to cluster these feature vectors. The Support Vector Machine (SVM) is utilized to train different recognition modules using the information of the clusters. Finally, the result of recognition is decided by each recognition module using the voting method.
The experimental image database contains the case of complex light changes, which includes two different sources of lights and the variations of light directions from fluorescent lamps and desk lamps. When the training images include all of light changes, the recognition rate of EigenFace is 94% and the proposed method can be up to 96%. If the testing images are with complex lighting changes than the training ones, the recognition rates of EigenFace and the proposed method are 31% and 63% , respectively. Obviously, the proposed method can increase the recognition rate by using the proposed facial components. In the training stage, the proposed method is more robust without special light changes images. Compared with EigenFace, the recognition rate of the proposed method reveals the great improvement. Experimental results show that the proposed method can indeed achieve reliable performance in face recognition.

[1] Z.H. Zhou and X. Geng, “Projection Functions for Eye Detection”, Pattern Recognition , vol. 37, pp.1049 - 1056, 2004.
[2] G.M. Beumer, Q. Tao, A.M. Bazen and R.N.J. Veldhuis, “A Landmark Paper in Face Recognition”, International Conference on Automatic Face and Gesture Recognition, pp.73–78, 2006.
[3] A. Bazen, R. Veldhuis and G. Croonen, “Likelihood Ratio-Based Detection of Facial Features”, 14th Annual Workshop on Circuits, Systems and Signal Processing, pp.323–329, 2003.
[4] P. Viola and M. Jones, “Robust Real-Time Object Detection”, International Journal of Computer Vision, 2002.
[5] L.I. Smith, “A Tutorial on Principal Components Analysis”, 2002.
[6] Y. Nara, J. Yang and Y. Suematsu, “Face Recognition Using Improved Principal Component Analysis”, Proc. International Symposium on Micromechatronics and Human Science, pp.77–82, 2003.
[7] H. Yu and J. Yang, “A Direct LDA Algorithm for High-Dimensional Data with Application to Face Recognition”, Pattern Recognition, vol. 34, no. 10, pp.2067–2070, 2001.
[8] H. Cevikalp, M. Neamtu, M. Wilkes and A. Barkana, “Discriminative Common Vectors for Face Recognition”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 27, no. 1, January 2005.
[9] J. Huang, P.C. Yuen, W.S. Chen and J.H. Lai, “Component-Based Subspace Linear Discriminant Analysis Method for Face Recognition with One Training Sample”, Optical Engineering, vol. 44, no. 5, May 2005.
[10] B. Heisele, P. Ho, J. Wu and T. Poggio, “Face Recognition: Component-Based Versus Global Approaches”, Computer Vision and Image Understanding, vol. 91, pp. 6–21, 2003.
[11] T. Moriyama, T. Kanade, J. Xiao and J.F. Cohn, “Meticulously Detailed Eye Region Model and Its Application to Analysis of Facial Images”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 5, May 2006.
[12] K. Jonsson, J. Matas, J. Kittler and Y. Li, “Learning Support Vectors for Face Verification and Recognition”, IEEE International Conference on Automatic Face and Gesture Recognition, pp.208–213, 2000.
[13] T.F. Cootes, G.J. Edwards and C.J. Taylor, “Active Appearance Models”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 23, no. 6, June 2001.
[14] G. Edwards, A. Lanitis, C. Taylor and T. Cootes, “Statistical Models of Face Images-Improving Specificity”, Image and Vision Computing, vol. 6, pp.203-211, 1998.
[15] J.T. Tou and R.C. Gonzalez, “Pattern Recognition Principles”, Addison-Wesley Publishing, 1974.
[16] J.C. Dunn, “A Fuzzy Relative of The ISODATA Process and Its Use in Detecting Compact Well-Separated Cluster”, Journal of Cybernetics, vol. 3, no. 3, pp.32–57, 1973.
[17] I. Rish, “An Empirical Study of The Naive Bayes Classifier”, Empirical Methods in Artificial Intelligence, 2001.
[18] K.C. Fukushima, “A Self-Organizing Multilayered Neural Network”, Biological Cybernetics, vol. 20, pp.121–136, 1975.
[19] L.E. Baum and T. Petrie, “Statistical Inference for Probabilistic Functions of Finite State Markov Chains”, Ann. Math. Stat., vol. 37, pp.1554–1563, 1966.
[20] J.C. Platt, N. Cristianini and J. Shawe-Taylor, “Large Margin DAG’s for Multiclass Classification, “Advances in Neural Information Processing Systems”, Cambridge, MA: MIT Press., vol. 12, pp.547–553, 2000.