近年來網路與行動裝置迅速發展，人們每天獲得的資訊量越來越多，並且隨著智慧型手機的普遍性上升，許多廣告商透過二維條碼作為傳遞訊息的管道。其中QR Code(Quick Response Code)是一種目前被廣泛應用在日常生活中的二維條碼，QR Code能夠傳輸網址、產品編號、文字訊息…等等，扮演著現實與數位世界之間的橋樑。然而，QR Code黑白相間的外觀和應用的場景格格不入，對於人類視覺上是一個相當突兀的存在，例如：廣告海報上放置的QR Code佔用了主要內容的空間且破壞整體美學設計的結構。因此，近幾年有許多研究以使用浮水印直接將訊息隱藏於圖片中作為研究方向，希望能以較美觀的浮水印技術取代QR Code。

浮水印技術主要目的為在視覺品質與隱藏內容解碼穩定性之間取得平衡，必須達到在人眼無法意識到圖片有隱藏訊息的情況下同時能以影像處理技術解密含有浮水印的圖片，以應用於真偽辨識、秘密訊息傳遞或版權保護等領域。本論文提出了一套基於深度學習方式的浮水印加解密系統，並且加入了人類視覺感知的考量以降低浮水印之可見性，藉由於訓練過程中加入雜訊可見度函數(Noise Visibility Function)作為參考，使得產生的浮水印貼近於原始影像的高頻區域，避開人眼較易察覺變動的低頻區域，達到視覺不可見性。此外，不同於傳統的數位浮水印只能應用於數位影像上，我們透過數位模擬影像失真訓練浮水印網路，以此方法訓練之網路能夠抵抗印刷及相機拍攝過程中造成的影像失真，因此能夠應用於現實環境中。

With the rapid development of internet and mobile devices in recent years,people are getting more and more information every day. Besides,as the ubiquity of smartphones rises,many advertisers use 2D barcodes as a channel of message transmission. QR Code (Quick Response Code) is one of the 2D barcodes that is widely used in daily life. QR code can transmit URL,product ID,text message,etc. It plays an important role in bridging reality and digital world. However, QR Code's black and white patterns are incompatible with the scene which it is placed on. It is an obtrusive existence for human vision. For instance,QR codes placed on advertisement posters take up the space for the main content and break the whole aesthetic design architecture. Thus,there were many researches that set their main target as using watermark to embed messages in pictures and hoped that they can replace QR Code with a more visually pleasing watermark technique.

The main target of watermarking is to maintain the balance between visual quality and robustness of decoding. Watermark techniques must achieve the goal of making hidden message imperceptible human eye and decoding the watermarked image robustly by computer vision techniques at the same time to be applied to various domains such as authenticity identification,secret communication and copyright protection. This paper proposes a watermarking system based on deep learning and integrated with human visual perception. By adding noise visibility function (NVF) as guidance during the training process,we force the watermark pattern to fit the high-frequency area of the image and therefore avoid the low-frequency area in which human can detect changes easily and achieve the invisibility of watermark. Furthermore,unlike traditional digital watermark techniques which can only be applied to digital images,our method can also resist the distortions resulting from physical transmission(i.e.,printing and capturing with camera)by using digital image distortion simulation during training. As a result,our watermark can be applied in real environments to reach the objective of replacing QR Codes.

[1] C. Godard, O. Mac Aodha, and G. J. Brostow, “Unsupervised monocular depth estimation
with left-right consistency,” in Proceedings of the IEEE conference on computer
vision and pattern recognition, pp. 270–279, 2017.
[2] M. Tancik, B. Mildenhall, and R. Ng, “Stegastamp: Invisible hyperlinks in physical
photographs,” in Proceedings of the IEEE/CVF Conference on Computer Vision and
Pattern Recognition (CVPR), pp. 2117–2126, June 2020.
[3] S. Baluja, “Hiding images within images,” IEEE transactions on pattern analysis
and machine intelligence, vol. 42, pp. 1685–1697, 2019.
[4] C. Zhang, P. Benz, A. Karjauv, G. Sun, and I. S. Kweon, “Udh: Universal deep
hiding for steganography, watermarking, and light field messaging,” Advances in
Neural Information Processing Systems, vol. 33, pp. 10223–10234, 2020.
[5] E. Wengrowski and K. Dana, “Light field messaging with deep photographic
steganography,” in Proceedings of the IEEE/CVF Conference on Computer Vision
and Pattern Recognition (CVPR), pp. 1515–1524, June 2019.
[6] J. Zhu, R. Kaplan, J. Johnson, and L. Fei-Fei, “Hidden: Hiding data with deep networks,”
in Proceedings of the European Conference on Computer Vision (ECCV),
pp. 657–672, 2018.
[7] A. Erickson, K. Kim, G. Bruder, and G. F. Welch, “Exploring the limitations of
environment lighting on optical see-through head-mounted displays,” in Symposium
on Spatial User Interaction, pp. 1–8, 2020.
[8] T. A. ©BIRD STUDIO/ SHUEISHA, “Dragonball heroes.” https://
www.dragonballheroes.com.tw/.
[9] O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy,
A. Khosla, M. Bernstein, A. C. Berg, and L. Fei-Fei, “ImageNet Large Scale
Visual Recognition Challenge,” International Journal of Computer Vision (IJCV),
vol. 115, pp. 211–252, 2015.
[10] F. Yi, G. Zhai, and Z. Zhu, “A robust circular two-dimensional barcode and decoding
method,” in 2019 Picture Coding Symposium (PCS), pp. 1–5, IEEE, 2019.
[11] S. Xie and H.-Z. Tan, “Blur-readable two-dimensional barcode based on blurinvariant
shape and geometric features,” International Journal of Advanced Robotic
Systems, vol. 18, p. 1729881421999589, 2021.
[12] G. Jancke, “High capacity color barcode,” Microsoft Research (online), available
from (accessed 2011-11-27), 2010.
[13] M. E. V. Melgar, M. C. Farias, F. de Barros Vidal, and A. Zaghetto, “A high density
colored 2d-barcode: Cqr code-9,” in 2016 29th SIBGRAPI Conference on Graphics,
Patterns and Images (SIBGRAPI), pp. 329–334, IEEE, 2016.
[14] C. Chen, B. Zhou, and W. H. Mow, “Ra code: A robust and aesthetic code for
resolution-constrained applications,” IEEE Transactions on Circuits and Systems for
Video Technology, vol. 28, pp. 3300–3312, 2017.
[15] C. Chen, W. Huang, B. Zhou, C. Liu, and W. H. Mow, “Picode: A new pictureembedding
2d barcode,” ieee transactions on image processing, vol. 25, pp. 3444–
3458, 2016.
[16] S.-S. Lin, M.-C. Hu, C.-H. Lee, and T.-Y. Lee, “Efficient qr code beautification with
high quality visual content,” IEEE Transactions on Multimedia, vol. 17, pp. 1515–
1524, 2015.
[17] H. Su, J. Niu, X. Liu, Q. Li, J. Wan, M. Xu, and T. Ren, “Artcoder: An end-toend
method for generating scanning-robust stylized qr codes,” in Proceedings of
the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 2277–
2286, 2021.
[18] N. Subramanian, O. Elharrouss, S. Al-Maadeed, and A. Bouridane, “Image
steganography: A review of the recent advances,” IEEE Access, vol. 9, pp. 23409–
23423, 2021.
[19] I. Cox, J. Kilian, F. Leighton, and T. Shamoon, “Secure spread spectrum watermarking
for multimedia,” IEEE Transactions on Image Processing, vol. 6, pp. 1673–1687,
1997.
[20] A. Reed, T. Filler, K. Falkenstern, and Y. Bai, “Watermarking spot colors in packaging,”
in Media Watermarking, Security, and Forensics 2015, vol. 9409, pp. 46–58,
SPIE, 2015.
[21] Y. Huang, B. Niu, H. Guan, and S. Zhang, “Enhancing image watermarking with
adaptive embedding parameter and psnr guarantee,” IEEE Transactions on Multimedia,
vol. 21, pp. 2447–2460, 2019.
[22] Q. Ying, J. Lin, Z. Qian, H. Xu, and X. Zhang, “Robust digital watermarking for
color images in combined dft and dt-cwt domains,” Mathematical Biosciences and
Engineering, vol. 16, pp. 4788–4801, 2019.
[23] D. Bhowmik, M. Oakes, and C. Abhayaratne, “Visual attention-based image watermarking,”
IEEE Access, vol. 4, pp. 8002–8018, 2016.
[24] S. Baluja, “Hiding images in plain sight: Deep steganography,” Advances in neural
information processing systems, vol. 30, 2017.
[25] J. Mannos and D. Sakrison, “The effects of a visual fidelity criterion of the encoding
of images,” IEEE Transactions on Information Theory, vol. 20, pp. 525–536, 1974.
[26] S. Voloshynovskiy, A. Herrigel, N. Baumgaertner, and T. Pun, “A stochastic approach
to content adaptive digital image watermarking,” in Proceedings of the Third
International Workshop on Information Hiding, p. 211–236, Springer-Verlag, 1999.
[27] O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for
biomedical image segmentation,” in International Conference on Medical image
computing and computer-assisted intervention, pp. 234–241, Springer, 2015.
[28] M. Jaderberg, K. Simonyan, A. Zisserman, et al., “Spatial transformer networks,”
Advances in neural information processing systems, vol. 28, 2015.
[29] M. J. Huiskes and M. S. Lew, “The mir flickr retrieval evaluation,” in Proceedings of
the 1st ACM international conference on Multimedia information retrieval, pp. 39–
43, 2008.
[30] R. Shin and D. Song, “Jpeg-resistant adversarial images,” in NIPS 2017 Workshop
on Machine Learning and Computer Security, vol. 1, p. 8, 2017.
[31] R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable
effectiveness of deep features as a perceptual metric,” in Proceedings of the IEEE
conference on computer vision and pattern recognition, pp. 586–595, 2018.
[32] L. J. Cronbach, “Coefficient alpha and the internal structure of tests,” psychometrika,
vol. 16, pp. 297–334, 1951.