本篇論文對X. Geng et al. 所提的人臉影像重建與年齡估計方法加以改進與解決其潛在的一些問題。對於訓練樣本中每一aging pattern中的遺失影像，以結合內插影像與該年齡的平均影像之加權平均結果來填補，填補後的結果稱之為 “全填滿年齡變化模式” (full-filled aging pattern)。相對於X. Geng et al. 使用PCA，本篇論文利用2DPCA訓練求得一個轉換矩陣，用來重建aging pattern中的遺失影像與估計人臉影像之年齡。在實驗中我們試不同年齡間距與不同長度之aging patterns訓練轉換矩陣，探討不同的因素如何影響測試結果。

This paper proposes an age estimation method that improves the method proposed by Xin Geng et al. in some respects: (1). Time efficiency is improved by using 2DPCA instead of PCA. (2). To solve the problem of having infinitely many solutions inherent in the system for solving the projection vectors, we modified the coefficient matrix of the system so the unique solution can be easily evaluated. (3). The proposed method was tested in several fashions to analyze possible factors that cause some problems.

目錄
第一章 緒論	1
1.1 研究動機與目的	1
1.2 系統流程	3
1.3 章節組織	4
第二章 相關文獻回顧	5
第三章 PCA與2DPCA	6
第四章 年齡變化模式與年齡估計	8
4.1 特徵擷取	8
4.2 全填滿年齡變化模式	10
4.3 轉換矩陣訓練	12
4.4 年齡變化影像模擬與年齡估計	14
第五章 實驗結果與分析	16
第六章 結論與未來研究	21
參考文獻	22
附錄：英文論文	24

圖目錄
Figure 1 訓練流程圖	3
Figure 2 測試流程圖	4
Figure 3 (a)原圖 (b)原圖加上68個特徵點	9
Figure 4 將原影像臉部特徵轉換至平均影像	10
Figure 5 aging pattern vector向量化過程[2]	10
Figure 6 X. Geng et al. [2]求得之“全填滿年齡變化模式” (full-filled aging pattern)	12
Figure 7 本方法求得之“全填滿年齡變化模式” (full-filled aging pattern)	12
Figure 8 填補第α個與第β個位置之間的遺失影像	12
Figure 9 FG-NET Aging Database	16
Figure 10 某一個體18, 24, 30, 45, 48歲之原始影像	17
Figure 11 在Figure 10之個體年齡0, 3, 6, 9, 12, …, 39, 42, 45, 48歲之原本與重建影像	17

表目錄
Table 1 年齡估計測試一之錯誤率	18
Table 2 年齡估計測試二之錯誤率	19
Table 3 年齡估計測試三之錯誤率	20

[1] 	I. Jolliffe, “Principal Component Analysis,” 2nd ed. Springer-Verlag, 2002. 
[2] 	X. Geng, Z.-H. Zhou, and K. Smith-Miles “Automatic Age Estimation Based on Facial Aging Patterns,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 29, No. 12, pp. 2234 -2240, 2007.
[3] 	G. J. Edwards, A. Lanitis, and C. J. Cootes, “Statistical Face Models: Improving Specificity,” Image Vision Computing, Vol. 16, No. 3, pp. 203-211, 1998.
[4] 	J.-Da Txia and C.-L. Huang, “Age Estimation Using AAM and Local Facial Features,” International Conference on Intelligent Information Hiding and Multimedia Signal Processing pp. 885-888, 2009.
[5] 	J. Suo, F. Min, S. Zhu, S. Shan, and X. Chen, “A Multi-Resolution Dynamic Model for Face Aging Simulation, ” IEEE Conference on Computer Vision and Pattern Recognition, June 2007.
[6] 	J. Yang and D. Zhang, “Two-Dimensional PCA:A New Approach to Appearance-Based Face Representation and Recognition, ” IEEE Tranactions on Pattern Analysis and Machine Intelligence, Vol. 26, No. 1, January 2004.
[7] 	H. Han, C. Otto, and A. K. Jain, “Age Estimation from Face Images: Human vs. Machine Performance, ” International Conference on Biometrics, June 2013.
[8] 	G. Guo and X. Wang, “A Study on Human Age Estimation under Facial Expression Changes,” Computer Vision and Pattern Recognition (CVPR), pp. 2547-2553, 2012.
[9] 	G. Guo, Y. Fu, C. Dyer, and T. S. Huang, “Image-Based Human Age Estimation by Manifold Learning and Locally Adjusted Robust Regression,” IEEE Trans. Image Processing, Vol. 17, No. 7, pp. 1178-1188, 2008.
[10] 	“FG-NET Aging Database,” http://www.fgnet.rsunit.com/
[11] 	A. Midori Albert, Karl Ricanek Jr., and Eric Patterson, “A Review of the Literature on the Aging Adult Skull and Face: Implications for Forensic Science Research and Applications,” Forensic Science International Vol. 172, Issue 1, 2, pp. 1-9, October 2007.