本論文針對國畫中的字元做字元擷取的動作，並將取得的字元位置利用樣本比對方法，將影像修復回尚未有題字或是蓋印的原樣。國畫為歷史文件，為了讓國畫可以永久保存不受毀壞因此建立數位檔案，且可提供人們線上瀏覽使用；因有許多字體與印章大多是後人收藏時才題字或蓋印，所以我們除了取出國畫上的字體外，更進一步將有字體的部份修復回單純只有畫作的影像。
國畫中的字體並無一定位置，大多在整張影像的四周，而背景由山水到花鳥皆有，並無一定格式。我們提出以型態學的背景重建(morphological reconstruct- tion)取得與四邊相連的背景資訊，再將與原影像相減反白後，取得文字的影像。接著使用各種型態學運算(morphological operactions)及connected component analysis來找出字元區塊，最後使用樣本比對方法修補得到的字元區塊。我們所提出的方法，經過實驗大多可得到令人滿意的成果。

More and more Chinese paintings are digitalized and browsed by people on the Internet. This paper proposes a two-folded scheme to extract calligraphy and seals from Chinese painting images and inpaint the original images such that images contain painting only.
To accomplish the extraction task, a morphological operation, reconstruction by erosion, is used to remove most content of a painting (i.e. background). Consequently, background positioned in the center and connected to borders of the image will be removed. As a result, calligraphy and seals become the principal object and making extraction task much simplified. To accomplish the inpainting task, an adaptive sampling selection method is used for automatically choosing a sample most close to the damaged pixel for inpainting. The proposed algorithm has been executed on a set of Chinese paintings and proved its efficacy.

第一章 緒論 1
1.1 研究動機與目的 1
1.2 研究內容 3
1.3 論文架構 4
第二章 相關研究與理論基礎 5
2.1 字元擷取相關研究 5
2.2 影像修補相關研究 8
2.3 型態影像學相關運算 11
2.4 二值化 20
第三章 研究方法 22
3.1 背景重建及二值化 22
3.2 選取字元範圍 26
3.2.1 去除雜訊 26
3.2.2 定義字元區塊範圍 28
3.2.3 後處理 35
3.3 影像修補 39
3.3.1 影像修補 39
3.3.2 動態樣本選取 42
第四章 實驗結果與探討 46
4.1 實驗結果 46
4.2 字元擷取相關探討 50
4.3 影像修補相關探討 56
第五章 未來研究方向 61
參考文獻 72

[1] Sittisak Rodtook, Yuttapong Rangsanseri, “Adaptive Thresholding of Document Images Based on Laplacian Sign,” International Conference on Information Technology : Coding and Computing, April 02-04, 2001, pp.501-505
[2] Q. Yuan, C.L. Tan, “Text Extraction from Gray Scale Document Images Using Edge Information,” Document Analysis and Recognition, 2001. Proceedings. Sixth International Conference on 10-13 Sept. 2001 pp.302-306.
[3] Keechul Jung, Kwang In Kim, Anil K. Jain, “Text information extraction in images and video : a survey,” Pattern Recognition Volume: 37, Issue: 5, May, 2004, pp.977-997.
[4] Wang E. K., Minya Chen, “A new robust algorithm for video text extraction,” Pattern Recognition Volume: 36, Issue: 6, June, 2003, pp.1397-1406
[5] Kongqiao Wang, Jari A. Kangas, “Character location in scene images from digital camera,” Pattern Recognition Volume: 36, Issue: 10, October, 2003 pp.2287-2299.
[6] Yassin M. Y. Hasan, Lina J. Karam, “Morphological Text Extraction from Images,” IEEE Transactions on Image Processing, Volume: 9, Issue: 11, November, 2000 pp.1978-1983.
[7] Manuel M. Oliveira, Brian Bowen, Richard McKenna, Yu-Sung Chang, “ Fast Digital Image Inpainting,” Appeared in the Procee-dings of the International Conference on Visualization, Imaging and Image Processing (VIIP 2001), Marbella, Spain. Septermer 3-5, 2001.
[8] Shin T. K., Rong-Chi Chang, Liang-Chen Lu, Huan-Chi Huang, “Multi-layer inpainting on Chinese artwork,” Multimedia and Expo. 2004 IEEE International Conference on Volume 1, 27-30 June 2004, pp.21-24.
[9] Li-Yi Wei, Marc Levoy, “Fast Texture Synthesis using Tree-structured Vector Quantization,” ACM SIGGRAPH 2000, Computer Graphics Proceedings, Annual Conference Series, July 2000, pp.479-488.
[10] Soo-Chang Pei, Yi-Chong Zeng, Ching-Hua Chang, “Virtual Restoration of Ancient Chinese Paintings Using Color Contrast Enhancement and Lacuna Texture Synthesis,” IEEE Transactions on Image Processing, Vol. 13, NO. 3, March 2004, pp.416-429.
[11] P. Soille, “Morphological Image Analysis: Principles and Applications,” Springer, Alemanha, 1999.
[12] Otsu N. A, “Threshold selection method from gray-level histograms[J]”. IEEE Trans Systems, Man and Cybernetics, 1979, 9(1): 62-66.
[13] “近代百年中國繪畫”，日本和泉市久保物記念美術館，2000年10月8日