由於運用廣泛，近年來人臉相關研究受到極大重視，然而其中針對臉型分類的研究卻相對稀少，有鑑於此，本研究針對臉型分類進行研究，並發展出一個植基於模糊半正定嵌入(Fuzzified Semidefinite Embedding, FSDE)的臉型分類系統。此系統結合膚色偵測、五官定位與下巴點偵測以得到臉部輪廓的初步位置，為了得到精確的臉部輪廓，本研究採用膚色遮罩結合區域梯度資訊進行精確輪廓定位。將臉部輪廓的切線角度利用模糊半正定嵌入進行特徵擷取後，以支持向量機(Support Vector Machine, SVM)進行分類。實驗結果顯示本系統能達到85%的分類正確率，這意味著本研究所提出的方法具有實用潛力。

Recently, automatic face recognition has attracted much research interest because of its variety of applications. Nevertheless, human face-shape classification is rarely noticed. In this thesis, a novel system of the face-contour extraction and face shape classification is proposed. First, rough location of facial contour is located by combining skin detection, facial feature registration and chin point detection. Then the skin mask and local gradient information are utilized to accurately locate contour. The tangential angle of facial contour is calculated for feature representation, and then Fuzzified SemiDefinite Embedding (FSDE) algorithm is invoked to reduce feature dimensionality. Finally, the reduced feature is classified by support vector machine. The experimental results show 85% of successful classification rates can be achieved.

【摘要】 I
【ABSTRACT】 II
【誌謝】 III
【目錄】 IV
【表目錄】 VI
第1章 序論 1
1.1 動機與目的 1
1.2 論文架構 3
1.3 系統架構 4
第2章 文獻探討 6
2.1 6
第3章 研究方法 11
3.1 前處理 11
3.2 特徵擷取 11
3.2.1 膚色偵測與五官定位 11
3.2.2 臉頰點、下巴點與下巴脖子交界點偵測 14
3.2.3 精確臉部輪廓 16
3.2.4 切線角度 18
3.3 維度縮減 19
3.3.1 半正定嵌入 19
3.3.2 模糊半正定嵌入 23
第4章 資料分群與標記 26
4.1 分群數 26
4.2 資料標記 27
第5章 實驗結果 28
5.1 實驗平台 28
5.2 實驗一　K-MEANS分群 29
5.3 實驗二　特徵擷取 32
5.4 實驗三　分類正確率 35
第6章 結論與未來展望 38
6.1 結論 38
6.2 未來展望 38
【參考文獻】 39

[1]李秀蓮, 攝影化妝. 台北市: 揚智文化事業股份有限公司, 1995.
[2]Y. Wu, J. Qiu and L. Ma, “Measurement and Classification of Shanghai Female Face Shape Based on 3D Image Feature,” In International Congress on Image and Signal Processing (CISP 2010), pages 2588-2592, 2010
[3]L.Q. Du, P. Jia, Z.T. Zhou and D.W. Hu, “Human face shape classification method based on active shape model,” Journal of Computer Applications, vol. 29, no. 10, pp. 2710-2715, October 2009.
[4]X. Wang, X. Zhang and J.L. Yao, “Skin Detection under Complex Background,” International Conference on Mechatronic Science, Electric Engineering and Computer, (EMC 2011), pages 1985-1988, 2011.
[5]J.S. Lee, “A Visual Context-Awareness System For Computer Room Classes – An Automatic Roll-Call System,” International Journal of Innovative Computing, Information and Control, (ICIC 2012) Vol. 8, no 9, pp. 1-11, September 2012.
[6]K.Q. Weinberger and L.K. Saul, “Unsupervised learning of image manifold by semidefinite programming,” Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2004), pages 988-995,2004.
[7]K. Crammer and Y. Singer, “On the Algorithmic Implementation of Multiclass Kernel-based Vector Machines,” Journal of Machine Learning Research , pp.265-292, December 2001.
[8]J.B. Tenenbaum, “Mapping a Manifold of Perceptual Observations,” Advances in Neural Information Processing Systems, 10, 1998.
[9]R. Pless and R. Souvenir, “A Survey of Manifold Learning for Images,” IPSJ Transactions on Computer Vision and Applications, vol. 1 pp.83-94, March 2009.
[10]J.B. MacQueen, “Some methods for classification and analysis of multivariate observations,” Proceedings of the Fifth Symposium on Math, Statistics, and Probability, pp.281-297, Berkeley, CA: University of California Press.
[11]L. Vandenberghe and S.P. Boyd, “Semidefinite Programming,” SIAM Review, vol.38, no.1, pp.49-95, March 1996.
[12]淺野八郎, 李玉瓊(譯), 人相術, 台北市: 大展出版社有限公司, 1998
[13]J. B. Tenenbaum, V. de Silva, and J. C. Langford. “A global geometric framework for nonlinear dimensionality reduction,” Science, vol.290, pp.2319–2323, December 2000.