基于Retinex先验引导的低光照图像快速增强方法

何磊; 易遵辉; 谢永芳; 陈超洋; 卢明

doi:10.16383/j.aas.c230585

基于Retinex先验引导的低光照图像快速增强方法

doi: 10.16383/j.aas.c230585

何磊^1,,
易遵辉^1,,
谢永芳^2,,
陈超洋^1,,
卢明^1,

1.
湖南科技大学信息与电气工程学院湘潭 411100
2.
中南大学自动化学院长沙 410000

基金项目: 国家重点研发计划“政府间国际创新合作”重点专项(2019YFE0118700), 国家自然科学基金(62222306, 61973110, 62203164), 湖南省教育厅科学研究项目(22A0349, 21B0499)资助

详细信息

作者简介:
何磊：湖南科技大学信息与电气工程学院讲师. 2017年和2023年分别获得山东大学学士学位和中南大学博士学位. 主要研究方向为视觉检测, 图像处理和深度学习. E-mail: helei_xb@hnust.edu.cn

易遵辉：湖南科技大学信息与电气工程学院讲师. 2017年和2023年分别获得山东大学学士学位和中南大学博士学位. 主要研究方向为光学成像, 图像处理和视觉检测. 本文通信作者. E-mail: yizunhui@hnust.edu.cn

谢永芳：中南大学自动化学院教授. 1999年获得中南工业大学博士学位. 主要研究方向为分散控制与鲁棒控制, 过程控制, 工业大数据和知识自动化. E-mail: yfxie@csu.edu.cn

陈超洋：湖南科技大学信息与电气工程学院教授. 2014年获得华中科技大学博士学位. 主要研究方向为群机器人系统协同控制, 复杂网络研究. E-mail: ouzk@163.com

卢明：湖南科技大学信息与电气工程学院教授. 2014年获得中南大学博士学位. 主要研究方向为流程工业工况识别与智能优化控制, 机器视觉与智能机器人. E-mail: mlu@hnust.edu.cn

计量
- 文章访问数: 170
- HTML全文浏览量: 66
- 被引次数: 0
出版历程
- 收稿日期: 2023-09-19
- 录用日期: 2024-01-08
- 网络出版日期: 2024-03-08

Fast Enhancement Method for Low Light Images Guided by Retinex Prior

1.
School of Information and Electrical Engineering, Hunan University of Science and Technology, Xiangtan 411100
2.
School of Automation, Central South University, Changsha 410000

Funds: Supported by National Key Research and Development Program of China for International Scientific and Technological Cooperation Projects (2019YFE0118700), National Natural Science Foundation of China (62222306, 61973110, 62203164), and Scientific Research Fund of Hunan Provincial Education Department (22A0349, 21B0499)

More Information

Author Bio:
HE Lei　Lecturer at the School of Information and Electrical Engineering, Hunan University of Science and Technology. He received his bachelor degree from Shandong University in 2017 and Ph.D. degree from Central South University in 2023. His research interest covers vision-based measurement, image processing, and deep learning

YI Zun-Hui　Lecturer at the School of Information and Electrical Engineering, Hunan University of Science and Technology. He received his bachelor degree from Shandong University in 2017 and Ph.D. degree from Central South University in 2023. His research interest covers optical imaging, image processing, and vision-based measurement. Corresponding author of this paper

XIE Yong-Fang　Professor at the School of Automation, Central South University. He received his Ph.D. degree from Central South University in 1999. His research interest covers decentralized control and robust control, process control, industrial big data, and knowledge automation

CHEN Chao-Yang　Professor at the School of Information and Electrical Engineering, Hunan University of Science and Technology. He received his Ph.D. degree from Huazhong University of Science and Technology in 2014. His research interest covers cooperative control of swarm robot systems and complex network research

LU Ming　Professor at the School of Information and Electrical Engineering, Hunan University of Science and Technology. He received his Ph.D. degree from Central South University in 2014. His research interest covers working condition recognition and intelligent optimization control of process industry, machine vision and intelligent robot

摘要

摘要: 低光照图像增强旨在提高在低光照环境下所采集图像的视觉质量. 然而, 现有的低光照图像增强方法难以在计算效率与增强性能之间达到很好的平衡, 为此, 提出一种基于Retinex先验引导的低光照图像快速增强方法, 将Retinex模型与Gamma校正相结合, 快速输出具有对比度高、视觉效果好和低噪声的图像. 为获取具有良好光照的图像以引导确定与输入图像尺寸大小一致的Gamma校正图, 提出基于Retinex模型的先验图像生成方法. 针对所提先验图像生成方法在极低光照区域中存在颜色失真的问题, 提出一种基于融合的Gamma校正图估计方法, 采用反正切变换恢复极低光照区域的颜色和对比度, 以提升Gamma校正图在极低光照区域的增强性能. 为抑制输出图像的噪声, 考虑到完全平滑的Gamma校正图不会平滑细节纹理的特点, 提出基于域变换递归滤波的Gamma校正图优化方法, 降低输出图像噪声的同时保持颜色和对比度. 实验结果表明, 所提方法不仅在主客观图像质量评价上优于现有大多数主流算法, 而且在计算效率上具有十分显著的优势.
- 低光照图像增强 /
- Gamma校正图 /
- Retinex模型 /
- 噪声抑制
Abstract: Low light image enhancement aims to improve the visual quality of images captured in low light environment. However, existing low light image enhancement methods are difficult to achieve a good balance between computational efficiency and enhancement performance. Therefore, a fast low light image enhancement method guided by Retinex prior is proposed. The Retinex model is combined with Gamma correction to quickly output images with high contrast, good visual effects and low noise. To obtain images with good illumination and guide the determination of Gamma correction map with the same size as the input image, a prior image generation method based on Retinex model is proposed. To solve the problem of color distortion in extremely low illumination areas, a fusion-based Gamma correction map estimation method is proposed, which uses arctangent transform to restore the color and contrast of extremely low illumination areas to improve the enhancement performance of Gamma correction map in extremely low illumination areas. To suppress the noise of the output image, considering that the completely smooth Gamma correction map cannot smooth the detail texture, an optimization method of Gamma correction map based on domain transform recursive filtering is proposed, which can reduce the noise and preserve the color and contrast of the output image. The experimental results show that the proposed method not only outperforms most of the existing mainstream algorithms in subjective and objective image quality evaluation, but also has significant advantages in computational efficiency.
- Low light image enhancement /
- Gamma correction map /
- Retinex model /
- noise suppression
注释:

1) 1¹ https://sites.google.com/site/vonikakis/datasets

HTML全文

图 1 不同方法的低光照增强结果

Fig. 1 Low light enhancement results of different methods

下载: 全尺寸图片幻灯片

图 2 所提方法的算法流程框架

Fig. 2 The algorithm flow framework of the proposed method

下载: 全尺寸图片幻灯片

图 3 获取先验图像的示例

Fig. 3 Example of obtaining the prior image

下载: 全尺寸图片幻灯片

图 4 先验图像与增强图像之间优势和劣势的可视化

Fig. 4 Visualization of advantages and disadvantages between prior image and enhanced image

下载: 全尺寸图片幻灯片

图 5 不同$ \alpha $取值下的权重函数曲线

Fig. 5 Weight function curves with different values of $ \alpha $

下载: 全尺寸图片幻灯片

图 6 室内场景下不同方法的低光照图像增强结果

Fig. 6 Low light image enhancement results of different methods in indoor scene

下载: 全尺寸图片幻灯片

图 7 室外场景下不同方法的低光照图像增强结果

Fig. 7 Low light image enhancement results of different methods in outdoor scene

下载: 全尺寸图片幻灯片

图 8 黑夜环境下不同方法的低光照图像增强结果

Fig. 8 Low light image enhancement results of different methods under dark night environment

下载: 全尺寸图片幻灯片

图 9 针对真实高噪声低光照图像下不同方法的增强结果

Fig. 9 Enhancement results of different methods for real high-noise low light image

下载: 全尺寸图片幻灯片

图 10 所提方法不同模块的消融研究

Fig. 10 Ablation studies of different modules of the proposed method

下载: 全尺寸图片幻灯片

表 1 不同方法在不同低光照数据集中的CEIQ值

Table 1 CEIQ values of different methods in different low light datasets

	NPEA^[34]	LIME^[8]	MF^[9]	LECARM^[35]	STAR^[36]	SCI^[16]	所提方法
DCIM	3.399	3.358	3.145	3.109	3.162	3.185	3.063
Fusion	3.405	3.399	3.428	3.405	3.387	3.340	3.358
NPE	3.784	3.478	3.499	3.497	3.484	3.523	3.429
VV	3.485	3.512	3.342	3.321	3.316	3.345	3.258
LIME	3.426	3.490	3.181	3.184	3.046	3.193	2.990
Darkface	3.198	3.246	2.779	2.716	2.438	2.407	2.496
平均值	3.449	3.413	3.229	3.205	3.139	3.165	3.099

下载: 导出CSV

表 2 不同方法在不同低光照数据集中的LOE值

Table 2 LOE values of different methods in different low light datasets

	NPEA^[34]	LIME^[8]	MF^[9]	LECARM^[35]	STAR^[36]	SCI^[16]	所提方法
DCIM	753.2	857.5	782.2	770.0	590.9	777.6	581.0
Fusion	564.9	803.6	614.2	681.7	527.4	703.1	523.0
NPE	731.2	847.6	750.1	845.2	549.3	849.7	588.1
VV	563.3	799.3	517.3	596.5	369.3	619.8	347.6
LIME	540.9	608.8	661.0	687.4	537.9	678.1	557.1
Darkface	955.8	942.1	883.9	428.6	621.8	458.9	260.8
平均值	684.9	809.9	701.4	668.2	532.7	681.2	476.2

下载: 导出CSV

表 3 不同方法在不同低光照数据集中的NL值

Table 3 NL values of different methods in different low light datasets

	NPEA^[34]	LIME^[8]	MF^[9]	LECARM^[35]	STAR^[36]	SCI^[16]	所提方法
DCIM	1.156	1.707	0.578	0.485	0.491	1.099	0.358
Fusion	0.561	0.715	0.509	0.599	0.597	0.510	0.325
NPE	0.778	0.908	0.690	0.685	0.688	0.623	0.526
VV	0.764	0.836	0.549	0.637	0.663	0.539	0.459
LIME	1.036	1.074	0.872	0.733	0.722	0.744	0.619
Darkface	2.019	2.586	1.585	1.157	1.147	1.084	0.789
平均值	1.052	1.304	0.797	0.716	0.718	0.766	0.512

下载: 导出CSV

表 4 不同方法处理不同图像尺寸的运行时间 (s)

Table 4 Running time for different image sizes processed by different methods (s)

	NPEA^[34]	LIME^[8]	MF^[9]	LECARM^[35]	STAR^[36]	所提方法
480×320	3.71	0.120	0.264	0.101	4.86	0.096
640×480	7.19	0.273	0.335	0.214	10.12	0.161
960×720	16.25	0.527	0.516	0.392	21.41	0.325
1280×720	18.01	0.677	0.603	0.463	24.63	0.417
1920×1080	48.33	1.348	1.321	0.989	64.32	0.935

下载: 导出CSV

参考文献(39)

[1]	徐少平, 张贵珍, 林珍玉, 刘婷云, 李崇禧. 一种多图像局部结构化融合的低照度图像增强算法. 自动化学报, 2022, 48(12): 2981−2995 Xu Shao-Ping, Zhang Gui-Zhen, Lin Zhen-Yu, Liu Ting-Yun, Li Chong-Xi. A multi-image local structured fusion-based low-light image enhancement algorithm. Acta Automatica Sinica, 2022, 48(12): 2981−2995
[2]	鞠明, 李成, 高山, 穆举国, 毕笃彦. 基于向心自动波交叉皮质模型的非均匀光照图像增强. 自动化学报, 2011, 37(7): 800−810 Ju Ming, Li Cheng, Gao Shan, Mu Ju-Guo, Bi Du-Yan. Non uniform-lighting image enhancement based on centripetal autowave intersecting cortical model. Acta Automatica Sinica, 2011, 37(7): 800−810
[3]	Kim T, Paik J. Adaptive contrast enhancement using gain-controllable clipped histogram equalization. IEEE Transactions on Consumer Electronics, 2008, 54(4): 1803−1810 doi: 10.1109/TCE.2008.4711238
[4]	Arici T, Dikbas S, Altunbasak Y. A histogram modifcation framework and its application for image contrast enhancement. IEEE Transactions on Image Processing, 2009, 18(9): 1921−1935 doi: 10.1109/TIP.2009.2021548
[5]	Farid H. Blind inverse gamma correction. IEEE Transactions on Image Processing, 2001, 10(10): 1428−1433 doi: 10.1109/83.951529
[6]	Huang S C, Cheng F C, Chiu Y S. Efficient contrast enhancement using adaptive gamma correction with weighting distribution. IEEE Transactions on Image Processing, 2013, 22(3): 1032−1041
[7]	Land E H, McCann J J. Lightness and Retinex theory. Journal of the Optical Society of America, 1971, 61(1): 1−11 doi: 10.1364/JOSA.61.000001
[8]	Guo X J, Li Y, Ling H B. LIME: Low-light image enhancement via illumination map estimation. IEEE Transactions on Image Processing, 2017, 26(2): 982−993 doi: 10.1109/TIP.2016.2639450
[9]	Fu X Y, Zeng D L, Huang Y, Liao Y H, Ding X H, Paisley J. A fusion-based enhancing method for weakly illuminated images. Signal Processing, 2016, 129: 82−96 doi: 10.1016/j.sigpro.2016.05.031
[10]	Li M D, Liu J Y, Yang W H, Sun X Y, Guo Z M. Structure-revealing low-light image enhancement via robust Retinex model. IEEE Transactions on Image Processing, 2018, 27(6): 2828−2841 doi: 10.1109/TIP.2018.2810539
[11]	Li J J, Feng X M, Hua Z. Low-light image enhancement via progressive-recursive network. IEEE Transactions on Circuits and Systems for Video Technology, 2021, 31(11): 4227−4240 doi: 10.1109/TCSVT.2021.3049940
[12]	Lv F F, Li Y, Lu F. Attention guided low-light image enhancement with a large scale low-light simulation dataset. International Journal of Computer Vision, 2021, 129: 2175−2193 doi: 10.1007/s11263-021-01466-8
[13]	Lu K, Zhang L H. TBEFN: A two-branch exposure-fusion network for low-light image enhancement. IEEE Transactions on Multimedia, 2021, 23: 4093−4105
[14]	Guo C L, Li C Y, Guo J C, Loy C C, Hou J H, Kwong S, et al. Zero-reference deep curve estimation for low-light image enhancement. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, USA: IEEE, 2020. 1777−1786
[15]	Zhao Z J, Xiong B S, Wang L, Ou Q F, Yu L, Kuang F. RetinexDIP: A unified deep framework for low-light image enhancement. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(3): 1076−1088 doi: 10.1109/TCSVT.2021.3073371
[16]	Ma L, Ma T Y, Liu R S, Fan X, Luo Z X. Toward fast, flexible, and robust low-light image enhancement. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, USA: IEEE, 2022. 5627−5636
[17]	Xu H T, Zhai G T, Wu X L, Yang X K. Generalized equalization model for image enhancement. IEEE Transactions on Multimedia, 2014, 16(1): 68−82 doi: 10.1109/TMM.2013.2283453
[18]	Pisano E D, Zong S Q, Hemminger B M, DeLuca M, Johnston R E. Contrast limited adaptive histogram equalization image processing to improve the detection of simulated spiculations in dense mammograms. Journal of Digital Imaging, 1998, 11: Article No. 193 doi: 10.1007/BF03178082
[19]	Rahman S, Rahman M M, Abdullah-Al-Wadud M, Al-Quaderi G D, Shoyaib M. An adaptive gamma correction for image enhancement. EURASIP Journal on Image and Video Processing, 2016, 2016(1): Article No. 35
[20]	Yang K F, Li H, Kuang H L, Li C Y, Li Y J. An adaptive method for image dynamic range adjustment. IEEE Transactions on Circuits and Systems for Video Technology, 2019, 29(3): 640−652
[21]	Jeong I, Lee C. An optimization-based approach to gamma correction parameter estimation for low-light image enhancement. Multimedia Tools and Applications, 2021, 80: 18027−18042 doi: 10.1007/s11042-021-10614-8
[22]	Jobson D J, Rahman Z, Woodell G A. Properties and performance of a center/surround Retinex. IEEE Transactions on Image Processing, 1997, 6(3): 451−462
[23]	Jobson D J, Rahman Z, Woodell G A. A multi-scale Retinex for bridging the gap between color images and the human observation of scenes. IEEE Transactions on Image Processing, 1997, 6(7): 965−976 doi: 10.1109/83.597272
[24]	Dabov K, Foi A, Katkovnik V, Egiazarian K. Image denoising by sparse 3-D transform-domain collaborative filtering. IEEE Transactions on Image Processing, 2007, 16(8): 2080−2095 doi: 10.1109/TIP.2007.901238
[25]	Fu X Y, Zeng D L, Huang Y, Zhang X P, Ding X H. A weighted variational model for simultaneous reflectance and illumination estimation. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, USA: IEEE, 2016. 2782−2790
[26]	Ren W Q, Liu S F, Ma L, Xu Q Q, Xu X Y, Cao X C, et al. Low-light image enhancement via a deep hybrid network. IEEE Transactions on Image Processing, 2019, 28(9): 4364−4375 doi: 10.1109/TIP.2019.2910412
[27]	Lore K G, Akintayo A, Sarkar S. LLNet: A deep autoencoder approach to natural low-light image enhancement. Pattern Recognition, 2017, 61: 650−662 doi: 10.1016/j.patcog.2016.06.008
[28]	Lv F F, Lu F, Wu J H, Lim C. MBLLEN: Low-light image/video enhancement using CNNs. In: Proceedings of the 29th British Machine Vision Conference. Newcastle, UK: BMVA, 2018. 1−13
[29]	Li C Y, Guo J C, Porikli F, Pang Y W. LightenNet: A convolutional neural network for weakly illuminated image enhancement. Pattern Recognition Letters, 2018, 104: 15−22 doi: 10.1016/j.patrec.2018.01.010
[30]	Jiang Y F, Gong X Y, Liu D, Cheng Y, Fang C, Shen X H, et al. EnlightenGAN: Deep light enhancement without paired supervision. IEEE Transactions on Image Processing, 2021, 30: 2340−2349 doi: 10.1109/TIP.2021.3051462
[31]	Fan G D, Fan B, Gan M, Chen G Y, Chen C L P. Multiscale low-light image enhancement network with illumination constraint. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(11): 7403−7417 doi: 10.1109/TCSVT.2022.3186880
[32]	Lempitsky V, Vedaldi A, Ulyanov D. Deep image prior. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018. 9446−9454
[33]	Gastal E S L, Oliveira M M. Domain transform for edge-aware image and video processing. In: Proceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference. Vancouver, Canada: ACM, 2011. 1–12
[34]	Wang S H, Zheng J, Hu H M, Li B. Naturalness preserved enhancement algorithm for non-uniform illumination images. IEEE Transactions on Image Processing, 2013, 22(9): 3538−3548 doi: 10.1109/TIP.2013.2261309
[35]	Ren Y R, Ying Z Q, Li T H, Li G. LECARM: Low-light image enhancement using the camera response model. IEEE Transactions on Circuits and Systems for Video Technology, 2019, 29(4): 968−981 doi: 10.1109/TCSVT.2018.2828141
[36]	Xu J, Hou Y K, Ren D W, Liu L, Zhu F, Yu M Y, et al. STAR: A structure and texture aware Retinex model. IEEE Transactions on Image Processing, 2020, 29: 5022−5037 doi: 10.1109/TIP.2020.2974060
[37]	Yang W H, Yuan Y, Ren W Q, Liu J Y, Scheirer W J, Wang Z Y, et al. Advancing image understanding in poor visibility environments: A collective benchmark study. arXiv preprint arXiv: 1904.04474, 2019.
[38]	Yan J, Li J, Fu X. No-reference quality assessment of contrast-distorted images using contrast enhancement. arXiv preprint arXiv: 1904.08879, 2019.
[39]	Dong L, Zhou J T, Tang Y Y. Noise level estimation for natural images based on scale-invariant kurtosis and piecewise stationarity. IEEE Transactions on Image Processing, 2017, 26(2): 1017−1030 doi: 10.1109/TIP.2016.2639447