LBG 演算法是其一最知名之解決向量量化編碼簿產生問題的演算法。近年來，啟發式演算法也在解決向量量化的編碼簿產生問題上已經佔有一席之地，主要原因為啟發式演算法讓各個研究能在有限時間內找到需要的近似解。而啟發式演算法可以分為兩個部份，其一是單一解的啟發式演算法，另一個是多重解的啟發式演算法。而因為多重解的啟發式演算法具有較高的可能性找到更好的解，現今大部分的研究都應用它去解決向量量化的問題，但多重解的啟發式演算法在計算時間上卻遠大於單一解演算法（包含單一解啟發式演算法和LBG 演算法），而單一解演算法能在短時間內得到不錯的效果卻容易受到初始解影響最終結果。因此本篇論文著重於向量量化初始編碼簿產生之改進，並提出兩個改進的方法。第一個取兩個初始化方法作結合去產生更好的初始解，另一個則是結合第一個方法跟維度差和的方法去產生較佳的初始解。並在模擬實驗下，應用在四個演算法：Linde-Buzo-Gray (LBG)、禁忌搜尋(Tabu search)、模擬退火(Simulated annealing) 及粒子群聚最佳化(PSO) 演算法，在實驗測試下，大致皆有不錯的品質提昇。

Linde-Buzo-Gray algorithm (LBG) is one of the most well-known image compressionalgorithms for the codebook generation problem of vector quantization. But in recent years,metaheuristic algorithms have also become more and more important for solving the codebookgeneration problem in vector quantization because they are capable of providing an approximatesolution within a reasonable time. Metaheuristic algorithms can be divided into twocategories: single solution-based and population-based algorithms. The population-based algorithmshave a higher possibility to get a better result, so most recent researches rely on themto solve the codebook generation problem in vector quantization, but they take more time thansingle solution algorithms and LBG algorithm. However, single solution algorithms are usuallyextremely sensitive to the initial solution. Thus, this paper presents an enhanced method tosolve the initial codebook generation problem of vector quantization that leverages the strengthof Yedla’s enhanced method for k-means initialization and Vimala’s ordered codebook generationmethod to create a better initial solution for the codebook generation problem. Simulationresults show that the proposed method can provide a quality that is significantly better than theoriginal algorithms can provide.

i. 論文審定書
iii. 誌謝
iv. 摘要
v. Abstract
viii. List of Figures
ix. List of Tables
Chapter 1. [簡介......1]
1.1 [動機......2]
1.2 [論文貢獻......2]
1.3 [論文架構......2]
Chapter 2. [文獻探討......3]
2.1 [向量量化(vector quantization)......3]
2.1.1 [編碼程序......3]
2.1.2 [編碼簿產生......3]
2.1.3 [解碼程序......5]
2.2 [單一解演算法(LBG演算法及單一解啟發式演算法)......5]
2.2.1 [Linde-Buzo-Gray (LBG)演算法......5]
2.2.2 [單一解啟發式演算法(single solution-based algorithms)......6]
2.2.2.1 [禁忌搜尋演算法(tabu search algorithm)......6]
2.2.2.2 [模擬退火演算法(simulated annealing algorithm)......7]
2.3 [多重解啟發式演算法(population-based algorithm)......9]
2.3.1 [粒子群聚最佳化演算法(particle swarm optimization algorithm)......9]
2.4 [編碼簿初始化產生問題......11]
2.4.1 [K-means初始化改進方法(Enhanced method for k-means initialization)......11]
2.4.2 [順序編碼簿產生(Ordered codebook generation)......12]
2.5 [總結......12]
Chapter 3 [改進演算法用於向量量化初始化問題......13]
3.1 [演算法的設計概念......13]
3.2 [演算法的流程......14]
3.2.1 [DSICS演算法的流程......15]
3.2.2 [EDSICS演算法運用歐幾里得距離及曼哈頓距離更新編碼簿......17]
3.3 [範例......19]
Chapter 4 [實驗結果......24]
4.1 [執行環境、參數設定、資料集介紹......24]
4.2 [模擬結果......25]
4.2.1 [品質(Quality)......26]
4.2.2 [計算時間(Computation Time)......27]
4.3 [總結......27]
Chapter 5 [結論與未來展望......38]
5.1 [結論......38]
5.2 [未來展望......38]
Bibliography [39]

[1] A. Gersho and R. M. Gray, Vector Quantization and Signal Compression. Springer, 1992.
[2] B. J. Jain and K. Obermayer, “Graph quantization,” Computer Vision and Image Understanding,vol. 115, no. 7, pp. 946–961, 2011.
[3] T. Abdel-Galil, Y. G. Hegazy, M. M. Salama, and R. Bartnikas, “Fast match-based vectorquantization partial discharge pulse pattern recognition,” Instrumentation and Measurement,vol. 54, no. 1, pp. 3–9, 2005.
[4] M. Kugler and H. S. Lopes, “Using a chain of lvq neural networks for pattern recognitionof eeg signals related to intermittent photic-stimulation,” in VII Brazilian Symposium onNeural Networks, pp. 173–177, IEEE, 2002.
[5] A. Laha, N. R. Pal, and B. Chanda, “Design of vector quantizer for image compressionusing self-organizing feature map and surface fitting,” Image Processing, vol. 13, no. 10,pp. 1291–1303, 2004.
[6] A. A. Mohammed, R. Minhas, Q. J. Wu, and M. Sid-Ahmed, “An efficient fingerprintimage compression technique based on wave atoms decomposition and multistage vectorquantization,” Integrated Computer-Aided Engineering, vol. 17, no. 1, pp. 29–40, 2010.
[7] M. Y. Kim and W. B. Kleijn, “Klt-based adaptive classified vq of the speech signal,”Speech and Audio Processing, vol. 12, no. 3, pp. 277–289, 2004.
[8] M. A. Hossan and M. A. Gregory, “Speaker recognition utilizing distributed DCT-II basedMel frequency cepstral coefficients and fuzzy vector quantization,” International Journalof Speech Technology, vol. 16, no. 1, pp. 103–113, 2013.
[9] K. Kotani, C. Qiu, and T. Ohmi, “Face recognition using vector quantization histogrammethod,” in International Conference on Image Processing, vol. 2, pp. II–105, IEEE,2002.
[10] J. Li, H. Li, A. Wang, A. Yan, and D. An, “Research of vector quantization algorithm inface recognition system,” in Proceedings of the International Conference on ElectronicEngineering, Communication and Management, pp. 447–453, Springer, 2012.
[11] Y. Linde, A. Buzo, and R. M. Gray, “An algorithm for vector quantizer design,” IEEETransactions on Communications, vol. 28, no. 1, pp. 84–95, 1980.
[12] F. Glover, “Future paths for integer programming and links to artificial intelligence,” Computers& Operations Research, vol. 13, no. 5, pp. 533–549, 1986.
[13] S. Kirkpatrick, M. Vecchi, et al., “Optimization by simmulated annealing,” Science, vol.220, no. 4598, pp. 671–680, 1983.
[14] J. Kennedy and E. Russell, “Particle swarm optimization,” in International Conference onNeural Networks, vol. 4, pp. 1942–1948, IEEE, 1995.
[15] S. Vimala, “Techniques for generating initial codebook for vector quantization,” in InternationalConference on Electronics Computer Technology (ICECT), vol. 4, pp. 201–208,IEEE, 2011.
[16] M. Yedla, S. R. Pathakota, and T. Srinivasa, “Enhancing k-means clustering algorithmwith improved initial center,” International Journal of Computer Science and InformationTechnologies, vol. 1, no. 2, pp. 121–125, 2010.
[17] T.-C. Lu and C.-Y. Chang, “A survey of vq codebook generation,” Journal of InformationHiding and Multimedia Signal Processing, vol. 1, no. 3, pp. 190–203, 2010.
[18] W.-H. A. Wang, S.-C. Yang, and C.-L. Tung, “Codebook design for vector quantizationusing genetic algorithm,” International Journal of Electronic Business, vol. 3, no. 2, pp.83–89, 2005.
[19] S. Lloyd, “Least squares quantization in pcm,” Information Theory, vol. 28, no. 2, pp.129–137, 1982.
[20] J. B. MacQueen, “Some methods for classification and analysis of multivariate observations,”in Proceedings of the Fifth Symposium on Math, Statistics, and Probability, pp.281–297, Oakland, 1967.
[21] P. Frnti, J. Kivijrvi, and O. Nevalainen, “Tabu search algorithm for codebook generationin vector quantization,” Pattern Recognition, vol. 31, no. 8, pp. 1139–1148, 1998.
[22] M.-H. Horng, “Vector quantization using the firefly algorithm for image compression,”Expert Systems with Applications, vol. 39, no. 1, pp. 1078–1091, 2012.
[23] F. Glover, “Tabu search-part i,” ORSA Journal on Computing, vol. 1, no. 3, pp. 190–206,1989.
[24] C. S. Sung and H. W. Jin, “A tabu-search-based heuristic for clustering,” Pattern Recognition,vol. 33, no. 5, pp. 849–858, 2000.
[25] N. Metropolis, A.W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, “Equationof state calculations by fast computing machines,” The Journal of Chemical Physics, vol.21, no. 6, pp. 1087–1092, 1953.
[26] S. Z. Selim and K. Alsultan, “A simulated annealing algorithm for the clustering problem,”Pattern Recognition, vol. 24, no. 10, pp. 1003–1008, 1991.
[27] J. Vaisey and A. Gersho, “Simulated annealing and codebook design,” in InternationalConference on Acoustics, Speech, and Signal Processing (ICASSP-88), pp. 1176–1179,IEEE, 1988.
[28] J. Flanagan, D. Morrell, R. Frost, C. J. Read, and B. E. Nelson, “Vector quantizationcodebook generation using simulated annealing,” in International Conference on Acoustics,Speech, and Signal Processing (ICASSP-89), pp. 1759–1762, IEEE, 1989.
[29] C. W. Reynolds, “Flocks, herds and schools: A distributed behavioral model,” ACM SiggraphComputer Graphics, vol. 21, no. 4, pp. 25–34, 1987.
[30] F. Heppner and U. Grenander, “A stochastic nonlinear model for coordinated bird flocks.,”American Association for The Advancement of Science, 1990.
[31] H. Kai-Cheng, C. Chun-Hao, T. Chun-Wei, and C. Ming-Chao, “An enhanced initializationmethod for codebook generation,” in International Conference on ConsumerElectronics-Taiwan (ICCE-TW), IEEE, 2015.
[32] C.-W. Tsai, C.-F. Lin, M.-C. Chiang, and C.-S. Yang, “A time-efficient particle swarmoptimization-based codebook generation algorithm,” in IEEE Congress on EvolutionaryComputation, pp. 1–6, IEEE, 2010.