臺灣博碩士論文加值系統

一般在的真假臉辨識過程中，常需用到昂貴的儀器(如光譜儀)來做測量以及在測量之後需要較長的資訊演算時間，而基於此缺點，本文提出一量測時不需用到昂貴儀器且演算快速的真假人臉辨識方法。在光源已知條件下，儀器上，只要以色度計量測待測臉之三刺激值X,Y,Z，便可求得反射係數之重建係數[1][2][3][4][5]；求得重建係數後，帶入本文所提出的真假人臉辨識演算法，演算流程分兩個階段，第一階段利用群落分析之Mahalanobis distance分析待測人臉之重建係數，此階段實驗結果，在真人人臉的辨識率方面可達100%，而假人臉的辨識率因同色異譜的現象產生誤判；第二階段為了克服同色異譜現象所造成之假人臉誤判，我們以CIELUV色彩空間為色差評估標準，利用同色異譜材質在兩光源下之色差不同為辨識基礎，成功的改良了假人臉的誤判現象，提升了假人臉的辨識率到達95.65%。

In the processing of discrimination between real and synthetic human faces, generally, we utilize expensive instrument to do measurement and take a long time to perform mathematical calculations on date measured. Accordingly, we propose an algorithm of discrimination between real and synthetic human faces which reduces the processing time and need not an expensive instrument to do measurement. On the condition of light source known, we can obtain the tristimulus values X, Y, Z of human faces by luminance colorimeter then acquire the reconstruction coefficient of reflectance.[1][2][3][4][5] After that we will regard the reconstruction coefficient as input date of processing stream. There are two steps in the processing stream of discrimination. Firstly, we use a multivariate analysis skill “Mahalanobis distance” to deal with the of reconstruction coefficient. The result reveal that the discrimination ratio of real human faces reaches to 100%, nevertheless, the discrimination of synthetic human faces is inaccurate since the phenomenon of metamerism. As a result, to overcome the inaccuracy of the discrimination of synthetic human faces, we utilize CIE LUV color space as a standard of color difference and base on a general idea that there are different values of color difference under two light sources on metamerism material. The general idea improves the fault discrimination of synthetic human faces and increases its discrimination ratio to 95.65%.

中文摘要 I
Abstract II
誌謝 III
目錄 IV
圖目錄 V
表格目錄 VI
第一章 導論 1
1.1 人臉辨識簡介 1
1.2 研究動機及研究內容 1
第二章 基礎理論 3
2.1 光線與顏色 3
2.2 標準色度學系統 6
2.3 均等色度空間 20
2.4 主成份分析 25
第三章 反射係數之測量實作與應用 28
3.1 反射係數介紹 28
3.2 實驗環境與方法 29
3.3 訓練集真人臉之反射係數主成分分析 33
第四章 真假人臉辨識 35
4.1 真假人臉辨識流程圖 35
4.2 真人臉資料庫建立 36
4.3 機率密度函數與馬氏距離 40
4.4 四個基底重建反射係數 46
4.5 色差與同色異譜現象 51
第五章 結論與未來發展 59
5.1 結論 59
5.2 未來發展 60
參考文獻 61

[1]C. van Trigt, "Smoothest reflectance functions. I. Definition and main results,"
J. Opt. Soc. Am. A 7, 1891-1904(1990).
[2]C. van Trigt, "Smoothest reflectance functions. II. Complete results," J. Opt.
Soc. Am. A 7, 2208-2222 (1990).
[3]C. van Trigt, ‘‘Color video system with Illuminant independent properties,’’
international patent application PCT/NL94/00049 (February 28, 1994).
[4]C. van Trigt, ‘‘Estimating reflectance: how to discount the illuminant,’’ Die
Farbe 40, 9–24 (1994).
[5]C. van Trigt, ‘‘Metameric blacks and estimating reflectance,’’J. Opt. Soc. Am.
A 11, 1003–1024 (1994).
[6]Nathan Moroney , Mark D. Fairchild , Robert W.G..Hunt, Changjun Li , M. Ronnier Luo and Todd Newman,”The CIECAM02 Color Appearance Model”.
[7]R. Missaoui, M. Sarifuddin and J. Vaillancourt, “Similarity measures for efficient content-based image retrieval, ”IEE Proc.-Vis. Image Signal Process., Vol. 152, No. 6, December 2005.
[8]L. Lucchese and S.K.Mitra,”Color segmentation based on separate anisotropic diffusion of chromatic and achromatic channels,” IEE Proc.-Vis. Image Signal Process., Vol. 148, No. 3, June 2001.
[9] 陳鴻興、陳君彥譯, 基礎色彩再現工程, 第2-16~2-21頁, 全華科技圖書股份有限公司, 民國九十二年。
[10]Menahem Friedman and Abraham Kandel, “Introduction to pattern recognition“pp. 118-124
[11]Wyszecki and Stiles, ”Color Science”pp.55-56 and 155
[12]Hhien-Che Lee” Introduction to color imaging science”pp183
[13] Rui Huang, Vladimir Pavlovic, and Dimitris N. Metaxas” A Hybrid Face
Recognition Method using Markov Random Fields” 0-7695-2128-2/04 (C) 2004
IEEE
[14]P. Belhumeur, J. Hespanha, andD.Kriegman.Eigenfacesvs.fisherfaces:Recognition
using class specific linear projection.IEEE T-PAMI, 19(7), 1997
[15] Neng-Chung Huand Cheng-Yeu Huang” Face Recognition of Real Human and
Artificial Human”