臺灣博碩士論文加值系統

人臉識別是一項熱門的計算機技術研究領域，它屬於生物特徵識別技術，是對生物體本身的生物特徵來區別生物體個體。生物特徵識別研究技術包括臉、指紋、掌紋、虹膜、聲音、體型、個人習慣等，相應的技術就有人臉識別、指紋識別、虹膜識別、視網膜識別、語音識別、簽字識別等。
本論文使用VQ直方圖法，這是一種簡單又可靠的人臉識別方法，通過將影像切成小塊，匹配碼向量取得索引，統計的索引為直方圖有效的個人特徵，利用二維離散小波處理、低通濾波處理、最小強度減法和VQ處理，產生直方圖，研究發現，加入同一個人的直方圖產生平均直方圖，能提升系統的識別率，將影像的尺寸適度的縮小可以提升處理速度而不影響識別率。
結果顯示，對於ORL人臉資料庫，40個人，每人10張照片，共400張圖片，每張照片在不同時間、不同光照、不同表情、不同人臉細節（戴眼鏡或者不戴眼鏡）下採集，而所有的影像均在黑暗均勻的背景下採集，人像為正面豎直人臉（部分有輕微旋轉），本論文達成的平均人臉識別率為96.2%。

Face recognition is a popular field of computer technology research which belongs to biometric identification technology, the biological can distinguish different biological by their own biological characteristics. Biometric identification techniques include face, fingerprint, palm, iris, sound, body and personal habits. The corresponding technology is face recognition, fingerprint recognition, iris recognition, retinal recognition, speech recognition and signature recognition.
This thesis uses the vector quantization (VQ) histogram method which is a simple and reliable face recognition method, by cutting the image into small pieces, and matching code vector to get indexed, and the statistical index is the effective personal feature of the histogram. Using two-dimensional discrete wavelet transform (DWT) processing, low-pass filtering processing, minimum intensity subtraction and VQ processing produce histogram. The study found that adding the same person's histogram to produce an average histogram can improve the recognition rate, and the size of the image is reduced appropriately can increase the processing speed without affecting the recognition rate.
The result shows that the ORL face database, experimental result show an average recognition rate of 96.2% for 400 images of 40 persons (10 images per person), every image collect at different times, different lighting, different expressions, different face details (wearing glasses or not wearing glasses).

目次
摘要 i
Abstract ii
目次 iii
表目次 v
圖目次 vi
第一章 簡介 1
1.1 前言 1
1.2 人臉識別概述 1
1.3 VQ技術概述 1
1.4 研究動機 3
1.5 論文大綱 4
第二章 文獻回顧 5
第三章 影像處理的基本知識 10
3.1 直方圖 10
3.2 空間平滑濾波器 10
3.2.1 平均濾波器 12
3.2.2 高斯濾波器 14
3.3 銳利化濾波器 15
3.4 小波轉換 16
第四章 人臉前處理與特徵萃取模組 18
4.1 系統架構 18
4.2 前處理模型 19
4.2.1 影像尺寸 21
4.2.2 影像切塊 21
4.2.3 最小強度減法 22
4.2.4 自定義碼簿 23
4.2.5 直方圖特徵 24
4.3 平均直方圖 26
4.4 距離度量 26
第五章 實驗結果與分析 29
5.1 實驗平台與資料庫 29
5.2 效能評估 29
5.3 實驗結果 30
5.3.1 直方圖特徵值 30
5.3.2 濾波器分析 32
5.3.3 最小強度減少與銳化濾波器分析 34
5.3.4 LBG碼簿與自定義碼簿 35
5.3.5 篩選重要的特徵 36
第六章 結論與討論 38
6.1 結論 38
6.2 後續研究方向 38
參考文獻 39

[1]K. Kotani, Chen Qiu and T. Ohmi, “Face recognition using vector quantization histogram method,” Proceedings, International Conference on Image Processing, Vol. 2, pp. II-105 - II-108, December 2002.
[2]Tzu-Chuen Lu and Ching-Yun Chang, “A Survey of VQ Codebook Generation,” Journal of Information Hiding and Multimedia Signal Processing, Vol. 1, No. 3, July 2010.
[3]Qiu Chen, Koji Kotani, Feifei Lee and Tadahiro Ohmi, “A Codebook Design Method for Robust VQ-Based Face Recognition Algorithm,” J. Software Engineering & Applications, Vol. 3, pp. 119-124, February 2010.
[4]Ahmed Aldhahab, Taif Alobaidi and Wasfy B. Mikhael, “Efficient Facial Recognition Using Vector Quantization of 2D DWT Features,” Signals, Systems and Computers, 2016 50th Asilomar Conference on, pp. 439-443, March 2016.
[5]Qiu Chen, Koji Kotani, Feifei Lee and Tadahiro Ohmi, “Face Recognition Using Markov Stationary Features and Vector Quantization Histogram,” 2014 IEEE 17th International Conference on Computational Science and Engineering, pp. 1934-1938, January 2014.
[6]Tadahiro Ohmi, Koji Kotani, Qiu Chen and Feifei Lee, “Face Recognition Algorithm using Vector Quantization Codebook Space Information Processing,” Intelligent Automation and Soft Computing, Vol. 10, No. 2, pp. 129-142, 2004.
[7]Qiu Chen, Koji Kotani, Feifei Lee, and Tadahiro Ohmi, “SIFT Features Using Vector Quantization histogram for Image Retrieval,” International Journal of Computer and Electrical Engineering, Vol. 4, No. 5, October 2012.
[8]Po-Yuan Yang, Jinn-Tsong Tsai and Jyh-Horng Chou, “PCA-Based Fast Search Method Using PCA-LBG-Based VQ Codebook for Codebook Search,” IEEE Access, Vol. 4, pp. 1332-1344, March 2016.
[9]Feifei Lee, Koji Kotani, Qiu Chen and Tadahiro Ohmi, ” Face Recognition Using Adjacent Pixel Intensity Difference Quantization Histogram,” IJCSNS International Journal of Computer Science and Network Security, Vol. 9, No. 8, August 2009.
[10]Dr. H.B. Kekre and Ms. Tanuja K. Sarode, “Vector Quantized Codebook Optimization using K-Means,” International Journal on Computer Science and Engineering, Vol.1(3), pp. 283-290, November 2009.
[11]Feifei Lee, Koji Kotani, Qiu Chen, and Tadahiro Ohmi, “Face Recognition Algorithm Based on Adjacent Pixel Intensity Difference Quantization in Rectangular Coordinate Plane,” International Journal of Computer and Electrical Engineering, Vol. 4, No. 4, August 2012.
[12]Di Liu, Dong-mei Sun and Zheng-ding Qiu, “Bag-of-Words Vector Quantization Based Face Identification,” 2009 Second International Symposium on Electronic Commerce and Security, Vol. 2, pp. 29-33, October 2009.
[13]H. Skinnemoen, “A codebook design method for fast VQ search,” 1996 IEEE Digital Signal Processing Workshop Proceedings, pp. 283-286, August 1996.
[14]Yan Yan, Qiu Chen and Feifei Lee, “Improved Face recognition using extended vector quantization histogram features,” 2016 IEEE International Conference on Signal and Image Processing (ICSIP), pp. 90-95, March 2016.
[15]ORL Database, “AT&T laboratories cambridge database of faces,” http://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html, (April 1992-1994)