具新的高效能與高效率之增強型自組織特徵映射圖於影像壓縮問題之研發

Development of New Effective and Efficient Enhanced Self-Organizing Map for Solving Image Compression Problem

林于峻(Yu-Jiun Lin)

指導教授 ： 蔡正發

國立屏東科技大學/管理學院/資訊管理系所/碩士(2009年)

https://doi.org/10.6346/NPUST.2009.00187

摘要

自組織特徵映射圖(Self-Organizing Map, SOM)是一種天生的分群演算法，其更適合應用於產生向量量化法(Vector Quantization)所需之壓縮影像編碼簿，原因在於SOM具有能產生良好的壓縮影像品質之優點，但其應用不只限於如此。自1980年，Kohonen提出自組織特徵映射圖後，眾多應用如雨後春筍般蓬勃發展，如市場區隔、鏡頭自動對焦、影像的壓縮、資料分群、圖樣辨識…等應用，自組織特徵映射圖的廣泛應用，顯示其天生的可應用性高，故相當具有研究價值。原因在於其映射資料樣本的成效相當良好，但SOM亦有另人詬病的缺點，即其訓練神經元的時間複雜度甚高，故本論文提出虛擬訓練樣本的概念，成功降低訓練所需執行時間及迭代次數。

關鍵字

影像壓縮 ； 向量量化 ； 自組織特徵映射圖

並列摘要

A self-organizing map (SOM), i.e. a congenital clustering algorithm, has a high compression ratio and produces high-quality reconstructed images, making it very suitable for generating image compression codebooks. However, SOMs incur heavy computation particularly when using large numbers of training samples. Thus, to speed up training, this thesis presents an enhanced SOM (named LazySOM) involving a hybrid algorithm combining LBG, SOM and Fast SOM. The proposed algorithm has a low computation cost, enabling the use of SOM with large numbers of training patterns. Simulations are performed to measure two indicators, PSNR and time cost, of the proposed LazySOM.

並列關鍵字

image compression ； SOM ； vector quantization

參考文獻

[16] Su, M.C., Liu, T.K., Chang, H.T., “Improving the self-organizing feature map algorithm using an efficient initialization scheme,” Tamkang Journal of Science and Engineering, Vol. 5, No. 1, pp. 35–48, 2002.

[5] Gray, R.M., “Vector Quantization,” IEEE ASSP Magazine, Vol. 1, No. 2, pp. 4-29, 1984.

[6] Sharma G, Trussell H J. Digital Color Imaging. IEEE Trans. on Image Processing, Vol. 6, No. 7, pp. 901–932, 1997.

[7] Linde, Y., Buzo, A., Gray, R.M., “An algorithm for vector quantization design,” IEEE Transaction Commun., Vol. COM-28, pp. 84–95, 1980.

[8] Kohonen, T., “Self-organizing map,” Berlim, 1995.