一种适用于无人艇光视觉感知的轻量型残差堆叠低照度图像增强网络

刘婷, 张宇欣, 王国峰, 罗佩琪, 范云生

doi:10.16411/j.cnki.issn1006-7736.2024.02.006

大连海事大学学报 >

2024 , Vol. 50 >Issue 2: 53 - 66

DOI: https://doi.org/10.16411/j.cnki.issn1006-7736.2024.02.006

一种适用于无人艇光视觉感知的轻量型残差堆叠低照度图像增强网络

展开

(大连海事大学船舶电气工程学院，辽宁大连 116026)

刘婷* ( 1987 － ) ，女，博士，讲师，E-mail: liuting0910@dlmu.edu.cn

网络出版日期: 2023-11-28

基金资助

中国博士后科学基金资助项目( 2019M661076 )

收起

A lightweight stacked residual low-light image enhancement network for USV optical vision perception

Expand

(College of Marine Electrical Engineering, Dalian Maritime University, Dalian 116026, China)

Online published: 2023-11-28

Fold

摘要

针对无人艇在低照度环境中感知困难问题，提出了一种轻量型残差堆叠低照度图像增强网络。首先，在特征融合中引入金字塔多尺度池化，以更好地保留图像细节。其次，引入深度可分离卷积，减轻网络负担，提高图像处理速度。此外，设计了一种新的复合损失函数，引入颜色损失减少颜色失真。最后，采用LeakyReLU激活函数防止神经元死亡。大量实验结果表明，相比于Stacked Attention Residual Network (SARN)低照度图像增强网络，本文方法在提升图像质量的同时加快了图像处理的速度。其中，结构相似性和峰值信噪比分别提高了3.31%和2.08%，模型计算量、参数量和单张处理时间分别减小了81.88%、75%和43.02%。因此，本文所提的方法适用于水面无人艇的低照度图像增强，具有一定的应用价值。

关键词： 低照度图像增强; 卷积神经网络; 深度可分离卷积; 金字塔池化; 颜色损失

本文引用格式

刘婷, 张宇欣, 王国峰, 罗佩琪, 范云生 . 一种适用于无人艇光视觉感知的轻量型残差堆叠低照度图像增强网络[J]. 大连海事大学学报, 2024 , 50(2) : 53 -66 . DOI: 10.16411/j.cnki.issn1006-7736.2024.02.006

Abstract

A lightweight stacked residual low-light image enhancement network is proposed to overcome the difficulty in accurately sensing the environment under low-light conditions for Unmanned surface vessels (USV). Firstly, a pyramid multi-scale pooling is introduced into feature fusion to better preserve image details. Secondly, depthwise separable convolution is introduced to lighten the network and improve the image processing speed. Thirdly, a new composite loss function with color loss is designed to reduce color distortion. Finally, LeakyReLU activation function is used to prevent neuronal death. Results from extensive experiments demonstrate that our method can effectively improve image quality while speeding up image processing compared to Stacked Attention Residual Network (SARN). Specifically, SSIM (Structure Similarity Index Measure) and PSNR (Peak Signal-to-Noise Ratio) are improved by 3.31% and 2.08% and FLOPs (Floating point operations), params and processing time of single image are reduced by 81.88%, 75% and 43.02%. Therefore, the proposed method is suitable for low-light image enhancement of USV, and has certain application value.

Key words： low-light image enhancement; convolutional neural network; depthwise separable convolution; Pyramid pooling; color loss

参考文献

［1］LI C, GUO C, HAN L, et al. Low-light image and video enhancement using deep learning: a survey［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(12): 9396-9416.
［2］张煜, 康哲, 马杰,等. 面向内河航运安全监控的多尺度船舶图像目标识别方法［J］. 大连海事大学学报， 2022, 48(1): 62-72.
ZHANG Y, KANG Z, MA J, et al. Multi-scale ship image target recognition method for inland navigation safety monitoring［J］. Journal of Dalian Maritime University, 2022, 48(1): 62-72. (in Chinese)
［3］JAIN A K. Fundamentals of digital image processing［M］. ［S.l.］: Prentice-Hall, Inc., 1989.
［4］LAND E H, McCANN J J. Lightness and retinex theory［J］. Journal of the Optical Society of America, 1971, 61(1): 1-11.
［5］JOBSON D J, RAHMAN Z, WOODELL G A. Properties and performance of a center/surround retinex［J］. IEEE Transactions on Image Processing, 1997, 6(3): 451-462.
［6］JOBSON D J, RAHMAN Z, WOODELL G A. A multiscale retinex for bridging the gap between color images and the human observation of scenes［J］. IEEE Transactions on Image Processing, 1997, 6(7): 965-976.
［7］WEI C, WANG W J, YANG W H, et al. Deep retinex decomposition for low-light enhancement［C］//British Machine Vision Conference. ［S.l.:s.n.］, 2018, Corpus ID: 52008443.
［8］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.［S.l.:s.n.］, 2019: 1632-1640.
［9］HAI J, XUAN Z,HAN S C R, et al. R2rnet: low-light image enhancement via real-low to real-normal network［J］. Journal of Visual Communication and Image Representation, 2023, 90: 103712.
［10］张航瑛,王雪琦,王华英,等.基于明度分量的Retinex-Net图像增强改进方法［J］.物理学报,2022,71(11):107-115.
ZHANG H Y, WANG X Q, WANG H Y, et al. Advanced Retinex-Net image enhancement method based on value component processing ［J］. Acta Physica Sinica, 2022, 71(11): 110701. (in Chinese)
［11］WANG W J, WEI C, YANG W H, et al. Gladnet: low-light enhancement network with global awareness［C］//2018 13th IEEE International Conference on Automatic Face & Gesture Recognition. ［S.l.］: IEEE, 2018: 751-755.
［12］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.
［13］张成，潘明阳，高翊然，等. 融合暗通道先验与循环生成对抗网络的航海图像去雾模型［J］. 大连海事大学学报， 2022, 48(4): 84-93.
ZHANG C, PAN M Y, GAO Y R, et al. Maritime image dehazing model based on cycle generative adversarial network incorporating with dark channel prior［J］. Journal of Dalian Maritime University, 2022, 48(4): 84-93. (in Chinese)
［14］XU X T, LI J J, HUA Z, et al. Attention-based multi-channel feature fusion enhancement network to process low-light images［J］. IET Image Processing, 2022, 16(12): 3374-3393.
［15］CUI H S, LI J J, HUA Z, et al. Attention-guided multi-scale feature fusion network for low-light image enhancement［J］. Frontiers in Neurorobotics, 2022, 16: 837208.
［16］KO S, PARK J, CHAE B, et al. Learning lightweight low-light enhancement network using pseudo well-exposed images［J］. IEEE Signal Processing Letters, 2021, 29: 289-293.
［17］CHEN Y H, LI Q F, LU D X, et al. A novel underwater image enhancement using optimal composite backbone network［J］. Biomimetics, 2023, 8(3): 275.
［18］HAO X Y, LIU L X. DGC-UWnet: underwater image enhancement based on computation-efficient convolution and channel shuffle［J］. IET Image Processing, 2023, 17(7): 2158-2167.
［19］LIU S X, LONG W, HE L, et al. Retinex-based fast algorithm for low-light image enhancement［J］. Entropy, 2021, 23(6): 746.
［20］JIANG Z Q, LI H T, LIU L J, et al. A switched view of Retinex: deep self-regularized low-light image enhancement［J］. Neurocomputing, 2021, 454: 361-372.
［21］OH J G, HONG M C. Low-light image enhancement using hybrid deep-learning and mixed-norm loss functions［J］. Sensors, 2022, 22(18): 6904.
［22］闫晓阳,王华珂,侯兴松，等.YCbCr空间分治的双分支低照度图像增强网络［J］.中国图象图形学报,2023,28(11):3415-3427.
YAN X Y，WANG H K， HOU X S, et al. Dual-branch low-light image enhancement network via YCbCr space divide-and-conquer［J］ Journal of Image and Graphics, 2023, 28（11）：3415-3427. (in Chinese)
［23］刘译隆，李颖，秦凌宇. 基于亮通道的单幅海域浓雾图像去雾算法［J］. 大连海事大学学报， 2022, 48(3): 103-112.
LIU Y L, LI Y, QIN L Y. Defogging algorithm for single sea fog images based on bright channel［J］. Journal of Dalian Maritime University, 2022, 48(3): 103-112. (in Chinese)
［24］GARG A, PAN X W, DUNG L R. LiCENt: low-light image enhancement using the light channel of HSL［J］. IEEE Access, 2022, 10: 33547-33560.
［25］IGNATOV A, KOBYSHEV N, TIMOFTE R, et al. Dslr-quality photos on mobile devices with deep convolutional networks［C］//Proceedings of the IEEE International Conference on Computer Vision.［S.l.:s.n.］, 2017: 3277-3285.
［26］ZHANG X, WANGX. MARN: multi-scale attention retinex network for low-light image enhancement［J］. IEEE Access, 2021, 9: 50939-50948.[HJ*2/3]
［27］WEI X X, ZHANG X S, LI Y J. Sarn: a lightweight stacked attention residual network for low-light image enhancement［C］//2021 6th International Conference on Robotics and Automation Engineering (ICRAE).
［S.l.］: IEEE, 2021: 275-279.
［28］PARK J, WOO S, LEE J Y,et al. A simple and light-weight attention module for convolutional neural networks［J］. International Journal of Computer Vision, 2020, 128(4): 783-798.
［29］ZHAO H S, SHI J P, QI X J, et al. Pyramid scene parsing network［C］//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. ［S.l.:s.n.］, 2017: 2881-2890.
［30］HARJOSEPUTRO Y, YUDA I, DANUKUSUMOK P. MobileNets: efficient convolutional neural network for identification of protected birds［J］. International Journal on Advanced Science, Engineering and Information Technology, 2020, 10(6): 2290-2296.
［31］GUO Y H, LI Y D, WANG L Q, et al. Depthwise convolution is all you need for learning multiple visual domains［C］//Proceedings of the AAAI Conference on Artificial Intelligence. ［S.l.:s.n.］, 2019, 33(1): 8368-8375.
［32］SIFRE L, MALLAT S. Rigid-motion scattering for texture classification［J］. arXiv.org, 2014, Corpus ID: 196607.
［33］GUO X J, LI Y, LING H B. LIME: low-light image enhancement via illumination map estimation［J］. IEEE Transactions on Image Processing, 2016, 26(2): 982-993.
［34］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, 2011, 21(1): 80-93.
［35］LOH Y P, CHAN C S. Getting to know low-light images with the exclusively dark dataset［J］. Computer Vision and Image Understanding, 2019, 178: 30-42.
［36］ZHANG Q, NIE Y W, ZHENG W S. Dual illumination estimation for robust exposure correction［J］. Computer Graphics Forum, 2019, doi:10.1111/cgf.13833.
［37］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/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 1780-1789.

Options

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract

参考文献