在這篇論文中，我們提出一個針對風景照的影像合成方法，藉由學習特徵的方式讓使用者只要任意輸入一張背景優美的影像以及一張前景影像(如人、動物..等)即可自動得到一個最佳的影像合成位置推薦並自動合成。找出一個最好的影像合成位置並非件容易的事情，除了需要經過使用者的主觀美感評估，不同的影像資料特性亦須考量。雖然在攝影中有一些簡單的合成規則，但從我們的主觀實驗中顯示出合成規則並不適用於許多例子。
我們的方法主要基於學習合成特徵的架構出發，藉由從專業攝影的角度學習美感的方式我們可以找出有別於傳統合成規則的潛在合成知識與規則。我們首先將不同的圖片類型進行分群，再利用空間與幾何紋理特徵在不同的影像群中學習專家所點出的不同的最佳合成位置推薦的潛在規則模型，這樣的好處有以下幾點1)可以適用於不同使用者所輸入的不同類別影像2)學習不同類別的潛在合成知識與規則。
為了證明我們所提供的方法為具有美感的最佳影像合成推薦位置，我們在主觀美感評估當中顯示了我們所提出的方法具有不錯的結果。此研究有許多不同的應用，其一是在合成影像中的最佳美感合成位置推薦並自動合成影像，再者是在攝影的方面我們提供攝影一個最佳主角位置的推薦。
總上所述，我們提出了一個基於美感特徵學習的影像合成系統，利用學習不同影像群間的影像特徵，讓普通使用者與業餘使用者皆可以達到不錯且具專業等級的合成/攝影效果。

In this paper, we present a learning photo composition for landscape images. The system will recommend the best image composed location by learning aesthetic features when a user arbitrarily inputs the beauty background image and foreground images (refine by image matting). It is difficult to find the best image composition location. In addition to different characteristics of images, the aesthetic assessments of users' subjective appreciations are needed to be considered and addressed. Although some rules like composition rules in photography, there are subjective assessments that are not appreciated or appropriate for some images.
Our method is based on learning feature that we could find the implicitly shared knowledge and rules among the professional photographers that is not like composition rules. In this work, we classified all types of the images into different sub-groups, and then we used spatial and geometry texture features to learn the model in the different sub-group. There are some advantages as the following: 1) our system could be applied to different type of images by user inputting 2) it provided results with different implicit knowledge and rules in different sub-group.
We proposed the learning photo composition for landscape images to find best location which can be adapted for different images. To prove this, in subjective aesthetic assessment for user studies showed our proposed method has better scores. This research has different applications. One of the applications is find the best composition location for image composition, and the other application is that it can provide the best object placed in view rectangular for photography.
In summary, we proposed a learning system for image composition in which we provided general users and amateurs alike to make good composition images that are relatively professional by using the aesthetic feature system for image composition proposed in our work.

摘 要
Abstract
Content
Chapter 1 Introduction
Chapter 2 Related Work
2.1 Image Cropping and Re-composition
2.2 Image Aesthetic Assessment
2.3 Color Compatibility
Chapter 3 Proposed Method
3.1 Overview
3.2 Data Construction
3.3 Image Sub-groups
3.4 Learning Features
3.4.1 Image decomposition
3.4.2 Features
3.4.3 Learning and Prediction
3.4.4 Post processing
3.4.5 Composition Features
Chapter 4 Experiments and Discussion
4.1 Data Set and User Study
4.2 Subjective Quality Evaluation
4.2.1 User voting
4.2.2 Results of composition images
Chapter 5 Conclusion
References

[1]B. Ni, M. Xu, B. Cheng, M. Wang, S. Yan, Q. Tian, “Learning to Photograph: A Compositional Perspective,” IEEE Trans. on Multimedia, vol. 15, no. 5, pp.1138-1150, 2013.
[2]J. Yan, S. Lin., S. B. Kang, and X. Tang, “Learning the change for automatic image cropping,” In IEEE Conference on Computer Vision and Pattern Recognition, 971–978, 2013.
[3]H.-H. Su, T.-W. Chen, C.-C. Kao, W. H. Hsu, and S.-Y. C. Chien,“Preference-aware view recommendation system for scenic photos based on bag of aesthetics-preserving features,” IEEE Trans. on Multimedia, vol. PP, no. 99, 2012.
[4]L. Liu, R. Chen, L. Wolf, and D. Cohen-Or, “Optimizing photo composition,” Computer Graphics Forum, vol. 29, no. 2, pp. 469–478, 2010.
[5]J. Wang and M. F. Cohen, “Image and video matting: a survey,” Found. Trends. Comput. Graph. Vis., vol. 3, pp. 97–175, 2007.
[6]P. P´erez, M. Gangnet, and A. Blake, “Poisson image editing,” ACM Trans. Graph., vol. 22, pp. 313–318, 2003.
[7]F.-L. Zhang, M. Wang, and S.-M. Hu. “Aesthetic Image Enhancement by Dependence-Aware Object Re-Composition,” IEEE Trans. on Multimedia, 15,2013.
[8]S. Bhattacharya, R. Sukthankar, and M. Shah. “A framework for photo-quality assessment and enhancement based on visual aesthetics,” in ACM MM’10, 2010, pp. 271–280.
[9][Online].Available:http://www.photographymad.com/pages/view/10-top-photography-composition-rules
[10]M. Nishiyama, T. Okabe, Y. Sato, and I. Sato, “Sensation-based photo cropping,” in Proceedings of the 17th ACM international conference on Multimedia, 2009, pp. 669–672.
[11]W. Lin and C.-C. J. Kuo, “Perceptual visual quality metrics: A survey,” J. of Visual Communication and Image Representation, vol. 22, no. 4, pp. 297–312, 2011.
[12]Y. Ke, X. Tang, and F. Jing, “The design of high-level features for photo quality assessment,” in Proc. Int. Conf. Comput. Vis. Pattern Recogn., pp. 419–426, 2006.
[13]R. Datta, D. Joshi, J. Li, and J. Z. Wang, “Studying aesthetics in photographic images using a computational approach,” in Proc. ECCV, pp. 7–13, 2006.
[14]C. Li, A. C. Loui, and T. Chen, “Towards aesthetics: a photo quality assessment and photo selection system,” in ACM MM’10, 2010, pp. 827–830.
[15]E. Reinhard, M. Ashikhmin, B. Gooch, and P. Shirley, “Color transfer between images,” IEEE Comput. Graphics Applic., vol. 21, pp. 34–41, 2001.
[16]B.-Y. Wong, K.-T. Shih, C.-K. Liang, and H. Chen, “Single image realism assessment and recoloring by color compatibility,” IEEE Trans.Multimedia, vol. 14, no. 3, pp. 760–769, 2012.
[17]S. Xue, A. Agarwala, J. Dorsey, and H. Rushmeier, “Understanding and improving the realism of image composites,” ACM Trans. Graph., vol. 31, no. 4, pp. 84:1–84:10, 2012.
[18]N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” in Proc. Int. Conf. Comput. Vis. Pattern Recogn., pp.886–893, 2005.
[19]P. F. Felzenszwalb and D. P. Huttenlocher, “Efficient graph-based image segmentation,” International Journal of Computer Vision, vol. 59, pp.167–181, 2004.
[20]C.-C. Chang and C.-J. Lin. “LIBSVM: A library for support vector machines,” ACM Trans. Intel. Syst. and Tech., 2:27:1–27:27, 2011.