雙維度注意力機制之 超解析度生成對抗式網路： 生物辨識影像為例

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：40

、訪客IP：3.147.77.76

姓名

黃啟恩(Chi-En Huang) 查詢紙本館藏

畢業系所

資訊工程學系

論文名稱

雙維度注意力機制之超解析度生成對抗式網路：生物辨識影像為例
(Super Resolution Generative Adversarial Network Base on the Dual Dimension Attention Mechanism ( For the case of Biometric Image Super-Resolution ))

相關論文

★ 基於虹膜色彩空間的極端學習機的多類型頭痛分類	★ 以多分數加權融合方式進行虹膜影像品質檢定
★ 基於深度學習之工業用智慧型機器視覺系統：以文字定位與辨識為例	★ 基於深度學習的即時血壓估測演算法
★ 基於深度學習之工業用智慧型機器視覺系統:以焊點品質檢測為例	★ 基於pix2pix深度學習模型之條件式虹膜影像生成架構
★ 以核方法化的相關濾波器之物件追蹤方法實作眼動儀系統	★ 雷射都普勒血流原型機之驗證與校正
★ 以生成對抗式網路產生特定目的影像—以虹膜影像為例	★ 一種基於Faster R-CNN的快速虹膜切割演算法
★ 運用深度學習、支持向量機及教導學習型最佳化分類糖尿病視網膜病變症狀	★ 應用卷積神經網路的虹膜遮罩預估
★ Collaborative Drama-based EFL Learning with Mobile Technology Support in Familiar Context	★ 可用於自動訓練深度學習網路的網頁服務
★ 基於深度學習方法之高精確度瞳孔放大片偵測演算法	★ 基於CNN方法之真假人臉識別模型

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

[檢視]

[下載]

本電子論文使用權限為同意立即開放。
已達開放權限電子全文僅授權使用者為學術研究之目的，進行個人非營利性質之檢索、閱讀、列印。
請遵守中華民國著作權法之相關規定，切勿任意重製、散佈、改作、轉貼、播送，以免觸法。

摘要(中)

隨著生物辨識技術的發展趨於成熟，許多系統逐漸運用生物特徵來取代傳統密碼以達到認證功能；雖然此方案解決了密碼認證系統的問題(例如：易受暴力攻擊、密碼過長而難以記憶等)，另一方面，卻也不可避免造成了其他缺陷（例如：生物特徵的存續、不變性，以及辨識率等）。

虹膜辨識作為生物辨識技術的一環，具有虹膜特徵持久存在、高辨識率等優勢，另一方面，人臉辨識則得益於較低的設備成本，並保持一定程度的辨識率；因而兩者皆用於設置重要服務的認證系統。然而，生物辨識系統的準確率主要取決於高自由度的特徵模式，低解析度的生物資訊影像將會造成特徵細節的損失進而大幅地降低辨識率；因而傳統上採用成本較高的取樣設備以獲得高解析度的影像。鑒於近年物聯網與移動裝置的技術熱潮，將成本較高的影像取樣設備佈署於移動裝置上，勢必造成整體設備費用的高昂，若使用成本低廉的設備也未必能滿足各種基於影像解析度的需求，例如：感光元件的硬體條件、高解析度的影像造成的儲存負擔、傳輸高解析度影像所消耗的頻寬等。

有鑑於此，本研究欲提出生物資訊影像的超解析度技術，使得原始的低解析度輸入影像能轉變為超高解析度影像，並得益於清晰的細節資訊來大幅地提升整體系統的辨識效果，並間接地降低低端設備的建置成本。

摘要(英)

Since the biometric solution is able to avoid the security problems which the traditional method may be compromised, for example, the password may be compromised by brute force attack or phishing attack, many systems gradually use biometrics to replace traditional passwords to achieve authentication functions. However, other potential defects of biometric are inevitably caused, such as the immutability of physiological characteristics, and the lower recognition rate in unconditional environment.
As a part of biometric technology, iris recognition has the advantages of features persistency and high recognition rate. On the other hand, face recognition benefits from lower equipment costs and maintains a certain recognition rate. Therefore, both authentication systems are commonly used for authentication services in business or military access control. However, the accuracy of the biometric system mainly depends on the features of high-degree-of-freedom. The low-resolution image will cause the loss of feature details and greatly reduce the biometric recognition rate. Therefore, the traditional solution attempt to apply the high-cost sampling equipment to achieve higher resolution images.
With the technology of the Internet of Things (IoT) become mature, the number of the IoT device will be expected to grow to 50 billion around the world in 2030, and the various service is already prepared to deploy on it. According to the traditional solution of iris recognition, the deployment of expensive image acquisition sensors on mobile devices will inevitably result in high retail price. However, the low-cost equipments may not be able to meet various requirements based on image resolution, such as the additive photosensitive element to catch the depth of field, the additive storage to store the high-resolution image, and the high bandwidth to transmit the high-resolution image.
In order to deal with above mentioned issue, the super-resolution technology have been proposed to generate the super-resolution image of biometrics from the low-resolution input. Benefit from clear detailed information of output biometrics image, the overall recognition system will be greatly improved, and reduce the cost of building low-end equipment indirectly.

關鍵字(中)

★ 超解析度
★ 生成對抗式網路
★ 生物辨識

關鍵字(英)

★ super-resolution
★ generative adversarial network
★ biometric

論文目次

中文摘要 i
Abstract ii
致謝 iv
目錄 v
圖目錄 viii
表目錄 xi
一、緒論 1
1-1研究背景 1
1-2 研究動機與目的 3
1-3論文架構 5
二、文獻回顧 6
2-1生物辨識技術 6
2-1-1生物辨識的流程與種類 7
2-1-2虹膜辨識 10
2-1-3人臉辨識 12
2-2超解析度網路介紹 14
2-2-1 SR-CNN 14
2-2-2 ESPCN 16
2-2-3 VDSR 16
2-2-4 SRDenseNet 17
2-2-5 EDSR 19
2-2-6 Residual Dense Network 20
2-2-7 SR-GAN 21
2-2-8 ESRGAN 24
2-2-9 nESRGAN+ 27
2-2-10 MA_SRGAN 29
2-2-11 RCAN 31
2-2-12 RAM 32
三、方法說明 34
3-1 方法架構 34
3-1-1 注意力機制的引入 34
3-1-2 整體網路架構 35
3-2 注意力機制模組 37
3-2-1 通道域注意力模組 (CAM) 38
3-2-2 空間域注意力模組 (SAM) 40
3-3 網路整體損失函數 42
四、實驗成果 44
4-1 實驗內容 44
4-1-1 資料集介紹 44
4-1-2 實驗資料集的劃分 45
4-1-3 實驗流程 47
4-1-4 實驗訓練的細節 51
4-2 實驗評價框架 52
4-3 實驗結果 53
4-3-1 虹膜影像客觀評價 54
4-3-2 人臉影像客觀評價 58
4-3-3 虹膜影像主觀評價 61
4-3-4 人臉影像主觀評價 66
五、結論與未來展望 73
5-1 結論 73
5-2 未來展望 74
六、參考文獻 75

參考文獻

[1] Li Xu, Jimmy SJ Ren, Ce Liu and Jiaya JiaDeep. Convolutional Neural Network for Image Deconvolution. In NIPS, 2014
[2] Wenzhe Shi1, Jose Caballero1, Ferenc Huszar, Johannes Totz1, Andrew P. Aitken1,
Rob Bishop1, Daniel Rueckert1, Zehan Wang1. Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network. In CVPR, 2016.
[3] Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, Yoshua Bengio. Generative Adversarial Nets. In NIPS, 2014.
[4] Justin Johnson, Alexandre Alahi, and Li Fei-Fei, “Perceptual losses for real-time style transfer and superresolution,” in European Conference on Computer Vision, Springer, 2016, pp. 694–711.
[5] John Daugman. How Iris Recognition Works. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 14, NO. 1, JANUARY 2004.
[6] Lanitis, Andreas. (2010). A survey of the effects of aging on biometric identity verification. International Journal of Biometrics. 2. 10.1504/IJBM.2010.030415.
[7] Aadhaar Certification. https://www.iritech.com/indian-business/aadhaar-certification (22/01/2021).
[8] INTERPOL Uses Facial Recognition to Catch Fugitive Murderer. https://findbiometrics.com/interpol-facial-recognition-catch-fugitive-murderer-502124/ (22/01/2021).
[9] Facial recognition scanners are already at some US airports. Here′s what to know. https://www.usatoday.com/story/travel/airline-news/2019/08/16/biometric-airport-screening-facial-recognition-everything-you-need-know/1998749001/ (22/01/2021).
[10] M. S. Hosseini, B. N. Araabi, and H. Soltanian-Zadeh, “Pigment Melanin: Pattern for Iris Recognition,” IEEE Transactions on Instrumentation and Measurement, vol. 59, no. 4, pp. 792–804, 2010. 1, 2, 3.
[11] ISO/IEC 29794-6:2015 Information technology - Biometric sample quality part 6: Iris image data. https://www.iso.org/standard/54066.html (22/01/2021).
[12] Socolinsky, Diego A.; Selinger, Andrea (1 January 2004). "Thermal Face Recognition in an Operational Scenario". IEEE Computer Society. pp. 1012–1019 – via ACM Digital Library.
[13] ISO/IEC 19794-5:2011 Information technology - Biometric data interchange formats part 5: Face image data. https://www.iso.org/standard/50867.html (22/01/2021).
[14] Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang, “Learning a deep convolutional network for image super-resolution,” in European Conference on Computer Vision. Springer, 2014, pp. 184–199.
[15] Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang, “Image super-resolution using deep convolutional networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 38, no. 2, pp. 295–307, 2016.
[16] Jiwon Kim, Jung Kwon Lee, and Kyoung Mu Lee, “Accurate image super-resolution using very deep convolutional networks,” in Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, 2016, pp. 1646–1654.
[17] Tong Tong, Gen Li, Xiejie Liu, Qinquan Gao, “Image super-resolution using dense skip connections,” in Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 4799-4807.
[18] Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kilian Q Weinberger, “Densely connected convolutional networks,” in Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, 2017, pp. 4700–4708.
[19] K. He, X. Zhang, S. Ren and J. Sun, "Deep Residual Learning for Image Recognition," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, 2016, pp. 770-778, doi: 10.1109/CVPR.2016.90.
[20] B. Lim, S. Son, H. Kim, S. Nah and K. M. Lee, "Enhanced Deep Residual Networks for Single Image Super-Resolution," 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Honolulu, HI, 2017, pp. 1132-1140, doi: 10.1109/CVPRW.2017.151.
[21] Yulun Zhang, Yapeng Tian, Yu Kong, Bineng Zhong, Yun Fu, “Residual dense network for image super-resolution,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 2472-2481.
[22] Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Ca- ´ballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, et al., “Photo-realistic single image super-resolution using a generative adversarial network,” in Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, 2017, pp. 4681–4690.
[23] S. Liu and W. Deng, "Very deep convolutional neural network-based image classification using small training sample size," 2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR), Kuala Lumpur, Malaysia, 2015, pp. 730-734, doi: 10.1109/ACPR.2015.7486599.
[24] Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy, “Esrgan: Enhanced super-resolution generative adversarial networks,” in The European Conference on Computer Vision Workshops (ECCVW), September 2018.
[25] Alexia Jolicoeur-Martineau, “The relativistic discriminator: a key element missing from standard gan,” arXiv preprint arXiv:1807.00734, 2018.
[26] Nathanael Carraz Rakotonirina, Andry Rasoanaivo, “Esrgan+: Further improving enhanced super-resolution generative adversarial network,” in International Conference on Acoustics, Speech and Signal Processing, 2020.
[27] Chi-En Huang and Yung-Hui Li, “MASK ATTENTION-SRGAN IN IRIS. RECOGNITION,” in Conference on Computer Vision, Graphics, and Image Processing (CVGIP), 2020
[28] Yulun Zhang, Kunpeng Li, Kai Li, Lichen Wang, Bineng Zhong, Yun Fu, RCAN: Image super-resolution using very deep residual channel attention networks, Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 286-301.
[29] Kim, J., Choi, J., Cheon, M., & Lee, J. (2018). RAM: Residual Attention Module for Single Image Super-Resolution. ArXiv, abs/1811.12043.
[30] Woo S., Park J., Lee JY., Kweon I.S. (2018) CBAM: Convolutional Block Attention Module. In: Ferrari V., Hebert M., Sminchisescu C., Weiss Y. (eds) Computer Vision – ECCV 2018. ECCV 2018. Lecture Notes in Computer Science, vol 11211. Springer, Cham. https://doi.org/10.1007/978-3-030-01234-2_1.
[31] J. Long, E. Shelhamer and T. Darrell, "Fully convolutional networks for semantic segmentation," 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 3431-3440, doi: 10.1109/CVPR.2015.7298965.

指導教授

栗永徽(Yung-Hui Li)

審核日期

2021-8-11

推文