簡易檢索 / 詳目顯示

回結果列表
研究生： 王寶秀
Wang, Pao-Shiou
論文名稱： 一個應用於零參考低光源影像增強的曝光注意力網路
An Exposure Attention Network for Zero-Reference Low-light Image Enhancement
指導教授： 戴顯權
Tai, Shen-Chuan
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 電機工程學系
Department of Electrical Engineering
論文出版年： 2024
畢業學年度： 112
語文別： 英文
論文頁數： 60
中文關鍵詞： 低光源影像增強 、曝光注意模塊 、深度學習
外文關鍵詞： low-light image enhancement, Exposure Attention Module, deep-learning
相關次數： 點閱：41  下載：1 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 在低光照條件下拍攝的圖像，常常給人眼帶來不佳的視覺體驗，同時也會影響計算機識別任務的效能。儘管深度學習技術在複雜視覺任務中取得了顯著進展，低光照圖像處理仍然有許多待解決的難題。主要面臨的三大挑戰：成對數據的獲取困難、深度學習所需的龐大計算資源，以及顏色偏差的問題。在此背景下，我們提出了一種輕量級的 U-Net架構，專為增強低光照圖像而設計。鑑於低光照圖像增強屬於低階任務，我們決定僅保留模型的淺層結構。受 SK 模塊的啟發，我們開發了一個曝光注意模塊（Exposure Attention Module），以幫助模型有效地捕捉曝光不足區域的細節。我們的實驗結果顯示，該模型所需參數僅為原始 U-Net 的 0.79%，但其增強效果足可匹敵。

    Images captured in low-light conditions not only yield poor visual effects for the human eye but also impede the performance of computer recognition tasks. Despite deep learning's remarkable advancements in sophisticated visual tasks, low-light image processing still presents substantial opportunities for improvement. Three primary challenges include: 1) the difficulty of obtaining paired data, 2) the excessive computational resources required for deep learning, and 3) color discrepancies. In this Thesis, a lightweight U-Net is tailored to enhance low-light conditions. Specifically, acknowledging that low-light image enhancement is a low-level task, only the shallow layers of the net are retained. Furthermore, inspired by the SK module [2], an Exposure Attention Module has been developed to assist the model in effectively conveying details from underexposed areas of images. The experiments show that the model requires only 0.79% the number of parameters compared to the original U-Net, yet it delivers comparable results.

    Abstract ii Acknowledgments iii Contents iv Chapter 1 Introduction 1 Chapter 2 Background and Related Works 3 2.1 Previous Work 4 2.1.1 Traditional Methods 4 2.1.2 Deep-learning Methods 6 2.2 Prior Knowledge 8 2.2.1 HSV 8 2.2.2 Retinex Theory 8 2.2.3 U-Net 10 2.2.4 SK module 12 2.2.5 Light Wrapping 13 Chapter 3 The Proposed Algorithm 14 3.1 Proposed Network Architecture 14 3.1.1 Propose Method Framework 14 3.1.2 Enhancing U-Net with Lightweight Techniques 17 3.1.3 Exposure Attention Module 19 3.2 Loss Functions 23 3.2.1 Exposure Control Loss 23 3.2.2 Entropy Loss 23 3.2.3 Reflectance Consistency Loss 24 3.2.4 Spatial Structure Loss 24 3.2.5 Illumination Smoothness Loss 24 3.2.6 Total Loss 25 3.3 Postprocessing 25 Chapter 4 Experiment Results 26 4.1 Experimental Datasets 26 4.1.1 Training Dataset 26 4.1.2 Test Datasets 27 4.2 Parameters and Experimental Setting 28 4.2.1 Experimental environment 28 4.2.2 Hyperparameters 28 4.3 Metrics 29 4.4 Ablation Experimental Results 30 4.5 Experimental Results 33 4.5.1 Quantitative Results 33 4.5.2 Qualitative Results 34 Chapter 5 Conclusion and Future Work 48 5.1 Conclusion 48 5.1 Future Work 49 References 50

    [1] Ronneberger, O., Fischer, P. and Brox, T. 'U-Net: Convolutional Networks for Biomedical Image Segmentation', MICCAI, 2015.
    [2] Li, X., Wang, W., Hu, X. and Yang, J. 'Selective kernel networks', Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 510–519, 2019.
    [3] Zheng, S., Ma, Y., Pan, J., Lu, C. and Gupta, G. 'Low-Light Image and Video Enhancement: A Comprehensive Survey and Beyond', IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, 2022.
    [4] He, S., Peng, B., Dong, J. and Du, Y. 'Mask-ShadowNet: Toward Shadow Removal via Masked Adaptive Instance Normalization', IEEE Signal Processing Letters, vol. 28, 2021.
    [5] Zamir, S.W., Arora, A., Khan, S., Hayat, M., Khan, F.S., Yang, M.H. and Shao, L. 'Learning enriched features for real image restoration and enhancement', European Conference on Computer Vision, pp. 492–511. Springer, 2020.
    [6] Zamir, S.W., Arora, A., Khan, S., Hayat, M., Khan, F.S., Yang, M.H. and Shao, L. 'Learning enriched features for fast image restoration and enhancement', IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45(2), pp. 1934–1948, 2022.
    [7] Guo, J., Ma, J., García-Fernández, Á.F., Zhang, Y. and Liang, H. 'A survey on image enhancement for Low-light images', Heliyon, 2023.
    [8] Zuiderveld, K. 'Contrast limited adaptive histogram equalization', Graphics Gems IV, pp. 474-485, 1994.
    [9] Smith, A.R. 'Color gamut transform pairs', ACM SIGGRAPH Computer Graphics, vol. 12, 1978.
    [10] Land, E.H. 'The Retinex', American Scientist, vol. 52, 1964.
    [11] Jobson, D.J., Rahman, Z.U. and Woodell, G.A. 'Properties and performance of a center/surround retinex', IEEE Transactions on Image Processing, vol. 6(3), pp. 451-462, 1997.
    [12] Liu, C., Wu, F. and Wang, X. 'EFINet: Restoration for low-light images via enhancement-fusion iterative network', IEEE Transactions on Circuits and Systems for Video Technology, vol. 32(12), pp. 8486-8499, 2022.
    [13] Jiang, Y., Gong, X., Liu, D., Cheng, Y., Fang, C., Shen, X., Yang, J., Zhou, P. and Wang, Z. 'Enlightengan: Deep light enhancement without paired supervision', IEEE Transactions on Image Processing, vol. 30, pp. 2340-2349, 2021.
    [14] Guo, C., Li, C., Guo, J., Loy, C.C., Hou, J., Kwong, S. and Cong, R. 'Zero-reference deep curve estimation for low-light image enhancement', CVPR, 2020.
    [15] Li, C., Guo, C. and Loy, C.C. 'Learning to enhance low-light image via zero-reference deep curve estimation', IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022.
    [16] Liang, D., Li, L., Wei, M., Yang, S., Zhang, L., Yang, W., Du, Y. and Zhou, H. 'Semantically contrastive learning for low-light image enhancement', Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36(2), pp. 1555-1563, 2022.
    [17] Nguyen, H., Tran, D., Nguyen, K. and Nguyen, R. 'PSENet: Progressive Self-Enhancement Network for Unsupervised Extreme-Light Image Enhancement', Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2023.
    [18] Li, Y., Niu, Y., Xu, R. and Chen, Y. 'Zero-referenced low-light image enhancement with adaptive filter network', Engineering Applications of Artificial Intelligence, vol. 124, p. 106611, 2023.
    [19] Google Research. 'Background features in Google Meet, powered by Web ML', Google Research Blog, 2 May. Available at: https://research.google/blog/background-features-in-google-meet-powered-by-web-ml/ (Accessed: 7 June 2024), 2023.
    [20] Jiang, Z., Li, H., Liu, L., Men, A. and Wang, H. 'A switched view of Retinex: Deep self-regularized low-light image enhancement', Neurocomputing, vol. 454, pp. 361-372, 2021.
    [21] Dinh, B.D., Nguyen, T.T., Tran, T.T. and Pham, V.T. '1M parameters are enough? A lightweight CNN-based model for medical image segmentation', Proceedings of the 2023 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), pp. 1279-1284. IEEE, 2023.
    [22] He, S., Peng, B., Dong, J. and Du, Y. 'Mask-ShadowNet: Toward shadow removal via masked adaptive instance normalization', IEEE Signal Processing Letters, vol. 28, pp. 957-961, 2021.
    [23] Shi, W., Caballero, J., Huszár, F., Totz, J., Aitken, A.P., Bishop, R., Rueckert, D. and Wang, Z. 'Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network', Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1874-1883, 2016.
    [24] Ma, S., Pan, W., Li, N., Du, S., Liu, H., Xu, B., Xu, C. and Li, X. 'Low-Light Image Enhancement using Retinex-based Network with Attention Mechanism', International Journal of Advanced Computer Science & Applications, vol. 15(1), 2024.
    [25] Haouassi, S. and Wu, D. 'An Efficient Attentional Image Dehazing Deep Network Using Two Color Space (ADMC2-net)', Sensors, vol. 24(2), p. 687, 2024.
    [26] Chung, H. Y. 'Low-Light Image Enhancement Techniques', Master's Thesis, National Cheng Kung University, 2023.
    [27] Garg, A., Pan, X.W. and Dung, L.R. 'LiCENt: Low-light image enhancement using the light channel of HSL', IEEE Access, vol. 10, pp. 33547-33560, 2022.
    [28] Cai, J., Gu, S. and Zhang, L. 'Learning a deep single image contrast enhancer from multi-exposure images', IEEE Transactions on Image Processing, vol. 27(4), pp. 2049-2062, 2018.
    [29] Wei, C., Wang, W., Yang, W. and Liu, J. 'Deep retinex decomposition for low-light enhancement', arXiv preprint arXiv:1808.04560, 2018.
    [30] Lee, C., Lee, C. and Kim, C.S. 'Contrast enhancement based on layered difference representation', Proceedings of the 2012 19th IEEE International Conference on Image Processing, pp. 965-968. IEEE, 2012.

    無法下載圖示 校內：立即公開
    校外：2026-07-29公開
    電子論文尚未授權公開，紙本請查館藏目錄
    QR CODE