超解析度成像 (Super-Resolution) 是透過軟體或硬體的方法提高原始圖像的解析度，
讓低解析度的圖像放大之後仍能保有清晰的輪廓，我們想將此技術應用在漫畫中。隨著
智慧型手機及電腦的普及，相比於傳統的實體漫畫，現代人較習慣使用螢幕閱讀、欣賞
漫畫，而網路上可觸及到的數位漫畫通常經過一定倍率的壓縮，我們希望使用超解析度
的方法將低分辨率的漫畫還原為高分辨率的漫畫；同時我們希望在高解析度觀賞漫畫下
仍有良好的體驗，而不是僅用內插的方式將漫畫放大，讓網點失去本質上的意義。我們
會先使用深度學習的方法偵測漫畫的網點區域，並對各區域進行網點的分類，而後根據
不同的網點類別使用不同的模型將圖像進行放大，最後合併放大的結果，讓漫畫放大的
同時仍能保持網點原有的密度。

Super resolution refers to the process of applying software or hardware to an image
to enhance its original resolution so that low-resolution images can retain clear silhouette
even after being magnified, and the goal of this study is to apply such techniques towards
manga. With the popularization of smartphones and computers, people nowadays are
more likely to read and appreciate manga strips on screen rather than in their traditional
paper format. However, digital manga found on the Internet have usually undergone some
degree of compression, which is why we hope to utilize super resolution to restore lowresolution manga images into high-resolution images while at the same time still offering
quality reading experience under high resolution, instead of simply applying interpolation
to magnify the images, which would render manga screentone purposeless. First, we apply
deep learning to determine which areas on the manga are filled with screentone and classify
them according to screentone patterns; then, we magnify different screentone types using
different models and, finally, integrate the magnified results. In this way, the magnified
manga are able to retain the original density of their screentone.

[1] Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang. Image superresolution using deep convolutional networks. CoRR, abs/1501.00092, 2015.
[2] Jiwon Kim, Jung Kwon Lee, and Kyoung Mu Lee. Accurate image super-resolution
using very deep convolutional networks. In The IEEE Conference on Computer
Vision and Pattern Recognition (CVPR Oral), June 2016.
[3] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning
for image recognition. CoRR, abs/1512.03385, 2015.
[4] Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew P. Aitken,
Alykhan Tejani, Johannes Totz, Zehan Wang, and Wenzhe Shi. Photo-realistic
single image super-resolution using a generative adversarial network. CoRR, abs/
1609.04802, 2016.
[5] Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley,
Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets.
In Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, and K. Q. Weinberger,
editors, Advances in Neural Information Processing Systems 27, pages 2672–2680.
Curran Associates, Inc., 2014.
[6] Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Chen Change
Loy, Yu Qiao, and Xiaoou Tang. ESRGAN: enhanced super-resolution generative
adversarial networks. CoRR, abs/1809.00219, 2018.
[7] C. Yao, S. Hung, G. Li, I. Chen, R. Adhitya, and Y. Lai. Manga vectorization and
manipulation with procedural simple screentone. IEEE Transactions on Visualization and Computer Graphics, 23(2):1070–1084, 2017.
[8] Chengze Li, Xueting Liu, and Tien-Tsin Wong. Deep extraction of manga structural lines. ACM Transactions on Graphics (SIGGRAPH 2017 issue), 36(4):117:1–
117:12, July 2017.
[9] L. Lettry, M. Perdoch, K. Vanhoey, and L. Van Gool. Repeated pattern detection
using cnn activations. In 2017 IEEE Winter Conference on Applications of Computer
Vision (WACV), pages 47–55, 2017.
39[10] Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. CoRR, abs/1505.04597, 2015.