[1]朱文佳,伊 君,杨正阳,等.双条件扩散概率模型驱动的低光照图像增强方法[J].南京师大学报(自然科学版),2025,48(05):114-120.[doi:10.3969/j.issn.1001-4616.2025.05.013]
　Zhu Wenjia,Yi Jun,Yang Zhengyang,et al.A Low-Light Image Enhancement Method with Dual Conditional Diffusion Probabilistic Models[J].Journal of Nanjing Normal University(Natural Science Edition),2025,48(05):114-120.[doi:10.3969/j.issn.1001-4616.2025.05.013]
点击复制

双条件扩散概率模型驱动的低光照图像增强方法()

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

卷:
48
期数:
2025年05期
页码:
114-120
栏目:
计算机科学与技术
出版日期:
2025-10-20

文章信息/Info

Title:
A Low-Light Image Enhancement Method with Dual Conditional Diffusion Probabilistic Models
文章编号:
1001-4616(2025)05-0114-07
作者:
朱文佳1; 3; 伊 君2; 杨正阳3; 陈凤欣3; 余 烨3
(1.安徽百诚慧通科技股份有限公司,安徽 合肥 230001)
(2.黄冈师范学院计算机与人工智能学院,湖北 黄冈 438000)
(3.合肥工业大学计算机与信息学院,安徽 合肥 230601)
Author(s):
Zhu Wenjia1; 3; Yi Jun2; Yang Zhengyang3; Chen Fengxin3; Yu Ye3
(1.Anhui Baichenghuitong Technology Co., Ltd., Hefei 230001, China)
(2.School of Computer Science and Artificial Intelligence, Huanggang Normal University, Huanggang 438000, China)
(3.School of Computer and Information, Hefei University of Technology, Hefei 230601, China)
关键词:
扩散概率模型; 低光照图像增强; 虚拟环境构建; 生成模型
Keywords:
diffusion probabilistic model; low-light image enhancement; construction of virtual environments; generative model
分类号:
TP391.41
DOI:
10.3969/j.issn.1001-4616.2025.05.013
文献标志码:
A
摘要:
针对当前低光照图像增强算法生成图像存在色彩恢复不足、噪声和模糊现象等问题,提出一种双条件扩散概率模型驱动的低光照图像增强方法(low-light image enhancement method with dual conditional diffusion probabilistic model,LLDiffusion). 该方法以低光照图像和直方图均衡化图像作为扩散概率模型中反向过程的条件输入,以便充分利用低光照图像内的有效信息. 同时,引入双条件噪声预测器,利用其多尺度特征提取模块和时间残差融合模块,获得更具真实性的生成图像,提高图像生成质量. 在现有基准数据集LOL和VE-LOL上对所提方法进行测试,实验结果表明,LLDiffusion在PSNR、SSIM、FID指标上均取得良好的效果,基于此方法增强的图像具有更好的曝光控制、更少的噪声和伪影.
Abstract:
To address the issues of inadequate color restoration, noise, and blurriness in images generated by current low-light image enhancement algorithms, a dual conditional diffusion probabilistic model-driven method for low-light image enhancement(LLDiffusion)is proposed. LLDiffusion takes low-light images and histogram equalized images as conditional inputs in the reverse process of diffusion probabilistic model, fully leveraging the valid information within the low-light image. Additionally, a dual conditional noise predictor is introduced, which includes multi-scale feature extraction module and a temporal residual fusion module to produce more realistic generated images, thereby improving the overall image quality. LLDiffusion was tested on benchmark datasets, LOL and VE-LOL. The experimental results demonstrate that the LLDiffusion achieves best performance in terms of FID, PSNR and SSIM. The enhanced images based on our method exhibit better exposure control, less noise, and fewer artifacts.

参考文献/References:

[1]LI M D,LIU J Y,YANG W H,et al. Structure-revealing low-light image enhancement via robust retinex model[J]. IEEE transactions on image processing,2018,27(6):2828-2841.
[2]WEI C,WANG W J,YANG W H,et al. Deep retinex decomposition for low-light enhancement[C]//Proceedings of the British Machine Vision Conference. Newcastle:BMVA Press,2018:155-166.
[3]ZHANG Y H,ZHANG J W,GUO X J. Kindling the darkness:a practical low-light image enhancer[C]//Proceedings of the ACM International Conference on Multimedia. Nice:Association for Computing Machinery,2019:1632-1640.
[4]YU Y,CHEN F X,YU J,et al. LMT-GP:combined latent mean-teacher and gaussian process for semi-supervised low-light image enhancement[C]//European Conference on Computer Vision. Milan:Springer,2024:261-279.
[5]余烨,陈维笑,陈凤欣. 面向车型识别的夜间车辆图像增强网络RIC-NVNet[J]. 中国图像图形学报,2023,28(7):2054-2067.
[6]WANG Y F,WAN R J,YANG W H,et al. Low-light image enhancement with normalizing flow[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver:AAAI Press,2022:2604-2612.
[7]YI X P,XU H,ZHANG H,et al. Diff-retinex:rethinking low-light image enhancement with a generative diffusion model[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Paris:IEEE,2023:12302-12311.
[8]SHANG K,SHAO M W,WANG C,et al. Multi-domain multi-scale diffusion model for low-light image enhancement[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver:AAAI Press,2024,38(5):4722-4730.
[9]朱文佳,张婷,程秋花,等. 考虑背景失真的无参考视频质量评价方法[J]. 南京师大学报(自然科学版),2025,48(3):102-111.
[10]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.
[11]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.
[12]WANG R,ZHANG Q,FU C W,et al. Underexposed photo enhancement using deep illumination estimation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Long Beach:IEEE,2019:6842-6850.
[13]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 IEEE Conference on Computer Vision and Pattern Recognition. Virtual Event:IEEE,2020:1777-1786.
[14]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 IEEE Conference on Computer Vision and Pattern Recognition. Virtual Event:IEEE,2021:10556-10565.
[15]HU Q,GUO X J. Low-light image enhancement via breaking down the darkness[J]. International journal of computer vision,2022,131(1):48-66.
[16]MA L,MA T Y,LIU R S,et al. Toward fast,flexible,and robust low-light image enhancement[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New Orleans:IEEE,2022:5627-5636.
[17]WANG T,ZHANG K H,SHEN T R,et al. Ultra-high-definition low-light image enhancement:a benchmark and transformer-based method[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Washington,D.C.:AAAI Press,2023:295.
[18]SHAKIBANIA H,RAOUFI S,KHOTANLOU H. Cdan:convolutional dense attention-guided network for low-light image enhancement[J]. Digital signal processing,2025,156:104802.
[19]JIANG H,LUO A,FAN H Q,et al. Low-light image enhancement with wavelet-based diffusion models[J]. ACM transactions on graphics(TOG),2023,42(6):1-14.
[20]YIN Y Y,XU D,TAN C C,et al. Cle diffusion:controllable light enhancement diffusion model[C]//Proceedings of the 31st ACM International Conference on Multimedia. Ottawa:ACM,2023:8145-8156.

备注/Memo

备注/Memo:
收稿日期:2024-12-10.
基金项目:安徽省自然科学基金项目(2308085MF216).
通讯作者:伊君,博士,讲师,研究方向:计算机视觉、多媒体技术与三维重建. E-mail:junyi@hgnu.edu.cn
更新日期/Last Update: 2025-10-20