隨著近年來的技術發展，在電腦科學領域中使用機器學習成為了廣泛討論的議題。透過的對訓練資料擷取特徵學習，使用機器自動創建的技術在許多領域蓬勃發展，例如：文學、工業設計、音樂、圖像等。
本篇論文關注於自動創作生成卡通圖像，基於輸入訓練圖集的非監督式合成創作系統。創作圖像這類研究，有研究使用深度學習網路對輸入圖片進行分析，後依照使用者的設定改變風格，或是創作出新的圖像。然而，使用深度學習網路的設備價格較高，取得不易，且運算時間長。
本篇論文所提出的系統成本較低且運算快，使用非監督式學習的方式，擷取輸入訓練圖集的區域部位(Region)特徵來合成創作。我們使用圖像切割得方式取得輸入訓練圖集的區域部位，計算部位自身的特徵，並對每張輸入圖片建立一有向連通圖描述區域部位間的鄰近相連關係(Region Relationship Graph)。再依前述兩種特性對所有區域做分群，分群結束後針對統計出的各群特性對分群結果做優化修正(Clustering Refinement)，最後依分群結果合成創作出新的圖像。
由實驗結果顯示，本篇論文所提出的系統能有效地對輸入訓練圖集有效的切割區域，並能在Clustering Refinement後提供良好的分群結果。由於創造的圖像結果是主觀的，經過我們的評分實驗結果顯示，系統創作出來的圖像是有普通標準以上的分數。

During the past few years, the use of machine learning in the field of computer science has become a subject of extensive discussion. Through learning features from training data, creative intelligence that gives machines ability to creation has been developed in many areas such as literature, industrial design, music, and images.
This thesis focuses on automatic creation of cartoon images, based on the input training examples. Unsupervised learning methods are applied to analyze the input training images. We use image segmentation to obtain regions. Then characteristics of regions are calculated. For each input image, we establish a connected region relationship graph to describe the relationship between the adjacent regions. All the regions from example images are clustered according to their characteristics. After clustering, the clustering refinement step is designed using the statistics of each cluster. Finally, the new images are created according to the results of clustering.
The experimental results show that the system proposed in this thesis can effectively segment the area of the training examples and provide good clustering results after clustering refinement. As the result of the creation of the image is subjective, we perform subjective tests on different users to score our creation results. The results are satisfying and the proposed system is computationally fast.

[1] Liu, Chien-Hung, and Chuan-Kang Ting. 2015. “Music Pattern Mining for Chromosome Representation in Evolutionary Composition.” In IEEE Congress on Evolutionary Computation (CEC 2015), Sendai, Japan, 25–28 May 2015, 2145–2152.
[2] Mordvintsev, Alexander, Olah, Christopher and Tyka, Mike, “DeepDream - a code example for visualizing Neural Networks”; Google Research (2015)
[3] Mordvintsev, Alexander, Olah, Christopher and Tyka, Mike,”Inceptionism: Going Deeper into Neural Networks”; Google Research (2015)
[4] Mordvintsev, Alexander, Olah, Christopher and Tyka, Mike, ”Inceptionism: Going Deeper into Neural Networks”; Google Research (2015)
[5] Leon A. Gatys, Alexander S. Ecker and Matthias Bethge,” A Neural Algorithm of Artistic Style”; arXiv preprint arXiv:1508.06576 (2015)
[6] Leon A. Gatys, Alexander S. Ecker and Matthias Bethge,” Image Style Transfer Using Convolutional Neural Networks”; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 2414-2423
[7] Szegedy hristian, Liu Wei, Jia Yangqing, Sermanet Pierre, Reed Scott, Anguelov Dragomir, Erhan Dumitru, Vanhoucke Vincent and Rabinovich Andrew, “Going Deeper with Convolutions”; Computing Research Repository. arXiv:1409.4842 (2014)
[8] R. Kimmel and A.M. Bruckstein,”On regularized Laplacian zero crossings and other optimal edge integrators”; International Journal of Computer Vision, 53(3) pages 225–243(2003)
[9] Nock, R. and F. Nielsen, Statistical region merging. IEEE Transactions on pattern analysis and machine intelligence, 2004. 26(11): p. 1452-1458.
[10] Samet, H.; Tamminen, M. (1988). "Efficient Component Labeling of Images of Arbitrary Dimension Represented by Linear Bintrees". IEEE Transactions on Pattern Analysis and Machine Intelligence.
[11] Michael B. Dillencourt; Hannan Samet; Markku Tamminen (1992). "A general approach to connected-component labeling for arbitrary image representations". Journal of the ACM.
[12] Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, and Clifford Stein], Introduction to Algorithms, Second Edition. MIT Press and McGraw-Hill, 2001. ISBN 0-262-03293-7. Section 22.2: Breadth-first search, pp. 531–539.
[13] MacQueen, J. B. (1967). Some Methods for classification and Analysis of Multivariate Observations. Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability. 1. University of California Press. pp. 281–297.
[14] Arthur, D.; Vassilvitskii, S. (2007). "k-means++: the advantages of careful seeding" (PDF). Proceedings of the eighteenth annual ACM-SIAM symposium on Discrete algorithms. Society for Industrial and Applied Mathematics Philadelphia, PA, USA. pp. 1027–1035.
[15] Canny, J., A computational approach to edge detection. IEEE Transactions on pattern analysis and machine intelligence, 1986(6): p. 679-698.
[16] Dan Motzenbecker and Kyle Phillips, “Google AutoDraw”, https://www.autodraw.com/
[17] Wei-Ta Chu and Chia-Hsiang Yu, "Optimized Speech Balloon Placement for Automatic Comics Generation," Proceedings of ACM International Workshop on Interactive Multimedia on Mobile and Portable Devices, pp. 1-6, 2013.