改进零参考深度曲线估计的低光图像增强方法

doi:10.3778/j.issn.1002-8331.2406-0291

摘要/Abstract

摘要： 零参考深度曲线估计（zero-reference deep curve estimation，Zero-DCE）虽然整体调整了图像的亮度，但却忽略了区域曝光和噪声过大的问题。针对上述问题，提出了一种更有效的基于照明曲线估计的低光图像增强方法，解决了不同曝光程度下，增强后图像亮度不均匀、存在过度曝光区域的问题，改善了增强后图像噪声的问题。在原有的U-Net基础上引入上采样块和残差网络，利用上采样块将不同层次的特征图进行融合，使得网络能够保留更多的原始图像信息，然后通过残差连接确保特征在传递过程中不会丢失，以此来更大程度地保留原有特征。加入各向异性扩散，在减轻图像噪声的同时保留边缘和结构细节。最后在非参考损失中引入通道一致性损失，抑制过度曝光区域。在MEF数据集上进行了视觉比较实验，在DICM、LIME、LOL、MEF数据集上进行了定量和定性实验。实验结果表明，该方法的SSIM值和PSNR值均取得较好结果。

关键词: 图像增强, 零参考, 各向异性扩散, 低光, 通道一致性损失

Abstract: Zero-reference deep curve estimation (Zero-DCE) adjusts image brightness globally but neglects the problems of regional exposure and excessive noise. To address these problems, this paper proposes a more effective low-light image enhancement method based on illumination curve estimation, which resolves issues of uneven brightness and overexposed areas in enhanced images under varying exposure levels and improves noise reduction. The paper introduces up-sampling blocks and a residual network into the original U-Net, enabling the network to retain more original image information by fusing feature maps from different layers. Residual connections ensure that features are not lost during transmission, preserving the original characteristics. Additionally, anisotropic diffusion reduces image noise while preserving edges and structural details. Channel consistency loss is introduced in the non-reference loss to suppress overexposed areas. The paper conducts visual comparison experiments on the MEF dataset and quantitative and qualitative experiments on the DICM, LIME, LOL and MEF datasets. The experimental results demonstrate that the proposed method achieves superior SSIM and PSNR values.

Key words: image enhancement, zero-reference, anisotropic diffusion, low-light, channel consistency loss

时圆圆, 陈纯毅, 于海洋, 胡小娟, 辛原旭. 改进零参考深度曲线估计的低光图像增强方法[J]. 计算机工程与应用, 2025, 61(16): 257-264.

SHI Yuanyuan, CHEN Chunyi, YU Haiyang, HU Xiaojuan, XIN Yuanxu. Low-Light Image Enhancement Method for Improving Zero-Reference Deep Curve Estimation[J]. Computer Engineering and Applications, 2025, 61(16): 257-264.

参考文献

[1] KUANG H L, ZHANG X S, LI Y J, et al. Nighttime vehicle detection based on bio?inspired image enhancement and weighted score-level feature fusion[J]. IEEE Transactions on Intelligent Transportation Systems, 2017, 18(4): 927-936.
[2] DING Y H, FAN H H, XU M L, et al. Adaptive exploration for unsupervised person re-identification[J]. ACM Transactions on Multimedia Computing, Communications, and Applications, 2020, 16(1): 1-19.
[3] AINETTER S, FRAUNDORFER F. End-to-end trainable deep neural network for robotic grasp detection and semantic segmentation from RGB[C]//Proceedings of the 2021 IEEE International Conference on Robotics and Automation. Piscataway: IEEE, 2021: 13452-13458.
[4] ABDULLAH-AL-WADUD M, KABIR M H, DEWAN M A A, et al. A dynamic histogram equalization for image contrast enhancement[J]. IEEE Transactions on Consumer Electronics, 2007, 53(2): 593-600.
[5] YANG J, XU Y, YUE H, et al. Low‐light image enhancement based on Retinex decomposition and adaptive gamma correction[J]. IET Image Processing, 2021, 15(5): 1189-1202.
[6] PARK S, YU S, MOON B, et al. Low-light image enhancement using variational optimization-based retinex model[J]. IEEE Transactions on Consumer Electronics, 2017, 63(2): 178-184.
[7] LI Y Z, NIU Y Z, XU R, et al. Zero-referenced low-light image enhancement with adaptive filter network[J]. Engineering Applications of Artificial Intelligence, 2023, 124: 106611.
[8] LORE K G, AKINTAYO A, SARKAR S. LLNet: a deep autoencoder approach to natural low-light image enhancement[J]. Pattern Recognition, 2017, 61: 650-662.
[9] WEI C, WANG W J, YANG W H, et al. Deep Retinex decomposition for low-light enhancement[J]. arXiv:1808.04560, 2018.
[10] JIANG Y F, GONG X Y, LIU D, et al. EnlightenGAN: deep light enhancement without paired supervision[J]. IEEE Transactions on Image Processing, 2021, 30: 2340-2349.
[11] ZHANG Y H, ZHANG J W, GUO X J. Kindling the darkness: a practical low-light image enhancer[C]//Proceedings of the 27th ACM International Conference on Multimedia. New York: ACM, 2019: 1632-1640.
[12] ZHANG Y H, GUO X J, MA J Y, et al. Beyond brightening low-light images[J]. International Journal of Computer Vision, 2021, 129(4): 1013-1037.
[13] 彭大鑫, 甄彤, 李智慧. 低光照图像增强研究方法综述[J]. 计算机工程与应用, 2023, 59(18): 14-27.
PENG D X, ZHEN T, LI Z H. Survey of research methods for low light image enhancement[J]. Computer Engineering and Applications, 2023, 59(18): 14-27.
[14] GUO C L, LI C Y, GUO J C, et al. Zero-reference deep curve estimation for low-light image enhancement[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 1777-1786.
[15] LI C Y, GUO C L, LOY C C. Learning to enhance low-light image via zero-reference deep curve estimation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(8): 4225-4238.
[16] LIU R S, MA L, ZHANG J A, et al. Retinex-inspired unrolling with cooperative prior architecture search for low-light image enhancement[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 10556-10565.
[17] CHENG F S, ZHANG J. Low-light image enhancement method based on illumination curve adjustment[J]. Journal of Physics: Conference Series, 2023, 2650(1): 012022.
[18] CAI J R, GU S H, ZHANG L, et al. Learning a deep single image contrast enhancer from multi-exposure images[J]. IEEE Transactions on Image Processing, 2018, 27(4): 2049-2062.
[19] FEKRI-ERSHAD S, TAJERIPOUR F. Impulse-noise resistant color-texture classification approach using hybrid color local binary patterns and Kullback-Leibler divergence[J]. The Computer Journal, 2017, 60(11): 1633-1648.
[20] MARTINEZ A, HERNANDEZ E, OLAGUE M, et al. Analytical-heuristic modeling and optimization for low-light image enhancement[J]. arXiv:2412.07659, 2024.
[21] LEE C, LEE C, LEE Y Y, et al. Power-constrained contrast enhancement for emissive displays based on histogram equalization[J]. IEEE Transactions on Image Processing, 2012, 21(1): 80-93.
[22] GUO X J, LI Y, LING H B. LIME: low-light image enhancement via illumination map estimation[J]. IEEE Transactions on Image Processing, 2017, 26(2): 982-993.
[23] LEE C, LEE C, KIM C S. Contrast enhancement based on layered difference representation[C]//Proceedings of the 2012 19th IEEE International Conference on Image Processing. Piscataway: IEEE, 2012: 965-968.
[24] LI C Y, GUO J C, PORIKLI F, et al. LightenNet: a convolutional neural network for weakly illuminated image enhancement[J]. Pattern Recognition Letters, 2018, 104: 15-22.
[25] CHEN Y H, ZHU G, WANG X Q, et al. FMR-Net: a fast multi-scale residual network for low-light image enhancement[J]. Multimedia Systems, 2024, 30(2): 73.
[26] MA L, MA T Y, LIU R S, et al. Toward fast, flexible, and robust low-light image enhancement[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 5627-5636.