臺灣博碩士論文加值系統

近年來分群的概念漸漸應用在影像壓縮上，而分群最常見的方式為K-means分群。K-means分群是利用亂數選取方式，讓所有分群的資料達到收斂。利用K-means收斂之後的資料點，建立編碼簿(code book)取代原始影像達到影像壓縮。本文提出利用遺傳演算法所做的編碼簿與K-means所做編碼簿做比較，之後在遺傳演算中加上權重(W)用來比較權重高低對編碼字(code word)個數與影像品質(PSNR)的影響。
在過去影像只能完整影像方式呈現，無法依個人喜好選擇所須要的部分圖像。隨著科技發展和技術的提升影像切割研究提出許多方法去解決。本文提出一個新的影像切割法K-merging，運用mean shift觀念將中心位置經過多次疊代，使相同類別資料歸為同群不再變動，之後將類別不同但資料相同作合併減少分群。

The K-means algorithm had been widely applied to the study of image compression, in recently years, it is mainly used to design the codebook. In our study, a genetic algorithm is proposed to accomplish the purpose, with a parameter(w), controlling the clustering result in the algorithm. In our experiments, image compression base on the genetic algorithm outperforms that based on the K-means algorithm.
The mean shift method had been applied to perform the image segmentation. Since the method segments the image based on pixels. The computation complexity is relatively high. In this study, we propose a K-merging method to segment the image based on blocks of image. The image segmentation based on K-merging method outperforms that based on the mean shift method in our study.

摘 要……I
Abstract……II
致謝……III
目 錄……IV
圖目錄……V
第壹章 緒論……1
第一節 研究背景……1
第二節 研究動機……2
第三節 研究目的……3
第貳章 文獻探討……4
第一節 向量量化(Vector Quantization)介紹……4
第二節 K-means clustering演算法介紹……6
第三節 遺傳演算法介紹……7
第四節 Mean shift演算法介紹……9
第參章 本論文方法……10
第一節 影像壓縮 - 遺傳演算法編碼簿設計……10
第二節 影像切割 - K-merging 切割法……13
第肆章 實驗結果……19
第一節 影像壓縮……19
第二節 影像切割……28
第伍章 結論……42
參考文獻……43
個人簡歷……44

[1]K. Krishna, K. R. Ramakrishnan , and M. A. L. Thathachar, "Vector Quantizationusing Genetic K-Means Algortihm for Image Compression," InternationalConference on Information, Communications and Signal Processing(ICICS '97),Singapore, 9-12 Sep. 1997.
[2]K. Krishna and M. N. Murty, "Genetic K-means Algorithms," Communicated toIEEE Transactions on Systems, Man and Cybernetics.
[3]J. Makhoul, S. Roucos, and H. Gish, "Vector Quantization in Speech Coding,"Proceeding of the IEEE, Nov. 1985
[4]N. M. Nasrabadi and R. A. King, "Image Coding Using Vector Quantization: AReview", IEEE Transaction on Communication, pp. 957-971, Aug. 1988.
[5]W. H. Equitz , "A New vector quantization clustering algorithm,", IEEE
Transaction ASSP, vol.37, pp.1568-1575, Oct.1989
[6]Y. C. Hu, and C. C. Chang, "A progressive Codebook Training Algorithm forImage Vector Quantization," Proceeding of the Fifth Asia-Pacific Conference onCommunications and Fourth Optoelectronics and CommunicationsConference(APCC/OECC 99'), Beijing, China, 1999, vol. 2 , pp.936-939
[7]Gonzalez, Richard E. Woods, Digital imageprocessing, N.J. :Prentice-Hall, UpperSaddle River, 2002, 2nd ed.
[8]M. Srinivas and L.M. Patnaik, "Genetic algorithms: a survey, " Computer , Vol.27 , June 1994, pp. 17 -26
[9]Comaniciu, D., Meer, P.: Mean shift analysis and applications. Proc. IEEE Int.Conf. on Computer Vision, Greece (1999) 1197-1203.
[10]Comaniciu, D., Meer, P.: Mean shift: a robust approach toward feature space analysis.IEEE Trans. on PAMI (2002) 603-619.