臺灣博碩士論文加值系統

近年來，由於科技的進步，使得手機與相機越來越發達，也使得人們拍照的越來越容易，相片數量因而大量增加。大部分的人拍照只為記錄生活，但如何拍出美感就值得探討。專業攝影師利用他們藝術天分拍下的那一瞬間表達的情感，而且部分的攝影師會用他們專業的美學技術來去增強照片進而更加呈現當時的意境。如何讓電腦學依據模糊的美學定義來分析美學相片與美學修圖來幫助人們是一個有意義且困難的挑戰。
本篇論文提出了一個基於當代美學學習之影像增強系統，在增強相片之前，我們使用許多方法去分析相片中的元素，例如高斯拉普拉斯變換、RMS方法、顯著圖方法等等，我們共有16種特徵去偵測相片中美學程度。增強部分我們提出一個新的概念使用分群結合分類來去分析我們提取16種特徵，這些方法可以依據我們美學數據集產生出4種風格群，當一張相片被決策樹方法判斷是不好的照片時，可以用決策樹的特性知道哪種特徵需要加以修改。我們列出10種可以修改照片的強大方法，例如伽馬校正方法、高斯模糊方法等等。最後我們使用區間減半法讓所增強的數值加速近似決策樹建議的增強值。
實驗部分為判斷美學相片，經實驗結果顯示平均正確率為96.8%，增強相片部分每一群代表一種相片的風格，我們使用決策樹特性來加以步步增強或降低不同的特徵，其相片後製成果就跟專業攝影師修圖一樣完美。

In recent years, the rise of smartphones and digital cameras makes it easier to take photos and a mass amount of photos are spread on the Internet. Photographic aesthetics is some sort of art which is expressed by the professional photographers’ aesthetic sensibilities and emotion. Moreover, many professional photographers make adjustments to the photos in post, and let photos much become more beautiful and meet the conditions of photographic aesthetics rules. Enhancing the images followed by ambiguous photographic aesthetics become a big task for computer.
In this thesis, an automatically image enhancement based on the aesthetics images dataset from the internet is proposed. We used many method to analyze an image such as RMS method, Laplace of Gaussian method, saliency map method, Gabor filter method and so on. We can use above sixteen features extracted from image to judge an image is good or not. We present a new concept to enhance images by using cluster styles which are generated from X-means and CART decision tree. When an input image is judged as a bad image by CART decision tree, the reason can be traced back by the decision tree characteristic to know which features needs enhancement. We list ten features which can enhance image efficiently such as gamma correction, Gaussian blur and so on. We use Interval Halving method to approach the value which come from giving suggestion of a feature by CART decision tree based on contemporary aesthetics criteria.
In the experiments, we apply cluster and classification to our dataset, and the average of cluster’s accuracy is 96.8%. In the enhancement part, we use CART decision tree aesthetic suggestion which means some feature are not enough or some feature are too high that can enhance our image step by step. Then we can get differently image style result like professional photographers do.

中文摘要 i
Abstract ii
致謝 iv
Contents v
List of Figures vii
List of Tables xii
Chapter 1 Introduction 1
1.1 Overview 1
1.2 Motivation 2
1.3 System Description 3
1.4 Thesis organization 4
Chapter 2 Background and Related Works 5
2.1 Objective Concept of Photographic Aesthetics 5
2.2 Related Proposals of Photographic Aesthetics and Image Enhancement 6
Chapter 3 Feature Extraction 11
3.1 Color components 11
3.2 Harmony 13
3.3 Blur and Sharpness 14
3.4 Contrast 16
3.5 Simplicity 17
3.5.1 Simplicity definition 17
3.5.2 Saliency map 18
3.5.3 Combine simplicity and saliency map 19
3.6 Colorfulness 20
3.7 Depth of Field 21
3.8 Homogeneous Texture Descriptor 23
3.9 Other Features 25
Chapter 4 Photograph Quality Prediction and Image Enhancement with Contemporary Criteria 28
4.1 X-means Algorithm 28
4.1.1 Basic concept of X-means algorithm 28
4.1.2 Image clustering 31
4.2 CART Algorithm 33
4.2.1 Basic concept of CART decision tree 33
4.2.2 Pruning 34
4.2.3 The training set assignment 35
4.3 Feature Enhancement 37
4.3.1 Gamma correction 38
4.3.2 Gaussian blur 40
4.3.3 Sharpness 42
4.3.4 Contrast 43
4.3.5 Other enhancement features 44
4.4 Approximation method 46
4.5 Image enhancement with contemporary aesthetics criteria 47
Chapter 5 Experimental Results and Discussions 53
5.1 Experimental Setup 53
5.2 Results of Cluster and Decision Tree Classification 56
5.3 The Results of Image Enhancement 63
Chapter 6 Conclusions and Future Works 69
6.1 Conclusions 69
6.2 Future Works 71
References 72

[1]D. Guy. “Definition Composition (photographic aesthetics)” In website Photokonnexion [online] accessed on 7/22, http://www.photokonnexion. com/3269-2/
[2]R. Datta, D. Joshi, J. Li, and J. Z. Wang, “Studying aesthetics in photographic images using a computational approach,” in Proceedings of the European Conference on Computer Vision, Graz, Austria, pp. 288-301, 2006.
[3]I. Daubechies, Ten Lectures on Wavelets, Society for Industrial and Applied Mathematics, Philadelphia, Pennsylvania, vol. 61, pp. 198-202, 1992.
[4]V. Ciesielski, P. Barile, and K. Trist, “Finding image features associated with high aesthetic value by machine learning,” Lecture Notes in Computer Science, Springer, Berlin, Germany vol. 7834, pp. 47-58, 2013
[5]Y. Ke, X. Tang, and F. Jing, “The design of high-level features for photo quality assessment,” in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, New York, pp. 419-426, 2006.
[6]H. Tong, M. Li, H. Zhang, and C. Zhang, “Blur detection for digital images using wavelet transform,” in Proceedings of the IEEE International Conference on Multimedia and Expo, Taipei, Taiwan, pp. 17-20, 2004.
[7]Y. Luo and X. Tang, “Photo and video quality evaluation: Focusing on the subject,” in Proceedings of the European Conference on Computer Vision, Marseille, France, pp. 386-399, 2008.

[8]C.-H. Yeh, Y.-C. Ho, B. A. Barsky, and M. Ouhyoung, “Personalized photograph ranking and selection system,” in Proceedings of the International Conference on Multimedia, Firenze, Italy, pp. 211-220, 2010.
[9]C. S. Fahn and M. L. Wu. “An autonomous aesthetics-driven photographing instructor with personality prediction.” in Proceedings of the International Conferences on Computer Graphics, Visualization, Computer Vision, and Game Technology, Jakarta, Indonesia, pp. 13-19, 2013.
[10]Y. M. Ro, M. Kim, H. K. Kang, B. Manjunath, and J. Kim, “MPEG-7 homogeneous texture descriptor,” Journal of Electronics and Telecommunications Research Institute , vol. 23, no. 2, pp. 41-51, 2001.
[11]S. J. Hwang, A. Kapoor, and S. B. Kang, “Context-based automatic local image enhancement,” in Proceedings of the European Conference on Computer Vision, Firenze, Italy, pp. 569-582, 2012.
[12]J. Yan, S. Lin, S. B. Kang, and X. Tang, “A Learning-to-Rank Approach for Image Color Enhancement,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Columbus, Ohio, pp. 2987-2994, 2014.
[13]A. Torralba, “Contextual priming for object detection,” International Journal of Computer Vision, vol. 53, no. 2, pp. 169-191, 2003.
[14]J. Harel, C. Koch, and P. Perona, “Graph-based visual saliency,” Advances in Neural Information Processing Systems, MIT Press, Cambridge, Massachusetts, vol. 19, pp. 545-552, 2006.
[15]C. Liu, J. Yuen, A. Torralba, J. Sivic, and W. T. Freeman, “Sift flow: Dense correspondence across different scenes,” in Proceedings of the European Conference on Computer Vision, Marseille, France, pp. 28-42, 2008.
[16]J. H. Friedman, “Greedy function approximation: a gradient boosting machine,” Annals of Statistics, vol. 29, no. 5, pp. 1189-1232, 2001.
[17]D. Cohen-Or, O. Sorkine, R. Gal, T. Leyvand, and Y.-Q. Xu, “Color harmonization,” ACM Transactions on Graphics, vol. 25, no. 3 pp. 624-630, 2006.
[18]E. Peli, “Contrast in complex images,” Journal of the Optical Society of America A, vol. 7, no. 10, pp. 2032-2040, 1990.
[19]M. Cheng, N. J. Mitra, X. Huang, P. H. Torr, and S. Hu, “Global contrast based salient region detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no.3 , pp. 569-582, 2015.
[20]Y. Rubner, C. Tomasi, and L. J. Guibas, “The earth mover's distance as a metric for image retrieval,” International Journal of Computer Vision, vol. 40, no. 2, pp. 99-121, 2000.
[21]M. Baştan, H. Cam, U. Güdükbay, and Ö. Ulusoy, “BilVideo-7: An MPEG-7 compatible video retrieval system with integrated support for complex multimodal queries,” IEEE MultiMedia, vol. 17, no.3, pp. 62-73, 2010.
[22]S. Jeannin, “MPEG-7 visual part of eXperimentation model, version 9.0,” Presented at the ISO/IEC JTC1/SC29/WG11/N3914, the 55th Mpeg Meeting, Pisa, Italy, 2001.
[23]J. MacQueen, “Some methods for classification and analysis of multivariate observations,” in Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Statistical Laboratory of the University of California, Berkeley, California, vol. 1, pp. 281-297, 1967.
[24]D. Pelleg and A. W. Moore, “X-means: Extending K-means with Efficient Estimation of the Number of Clusters,” in Proceedings of the International Conference on Machine Learning, Stanford, California, pp. 727-734, 2000.
[25]L. Breiman, J. Friedman, C. J. Stone, and R. A. Olshen, Classification and Regression Trees, Wadsworth International Group, Belmont, California, 1984.
[26]X. Wu, V. Kumar, J. R. Quinlan, J. Ghosh, Q. Yang, H. Motoda, et al., “Top 10 algorithms in data mining,” Knowledge and Information Systems, vol. 14, no.1, pp. 1-37, 2008.