Underwater image enhancement based on degradation type awareness
-
摘要: 高质量的水下光学影像是海底场景数字孪生、底栖生境保护、海底矿产资源探测和海底未知现象理解等任务的重要依托。然而，受复杂水体环境、光照条件等因素的影响，水下光学影像存在颜色失真、细节模糊、对比度低等退化问题。现有水下图像增强方法多聚焦于增强算法本身的优化，缺乏对不同退化类型图像的溯源、分类与分级等系统性分析机制。为此，针对水下光学图像成像环境的复杂性和异质性，本文提出一种顾及退化类型的图像质量增强策略。首先，构建了水下图像退化类型感知网络用于识别水下雾化模糊图像，准确率达到97%，对光照退化类型亦具有较高的区分能力。其次，针对识别出的水下雾化图像，基于真实水下图像色偏值的分布统计，设计自适应颜色校正方法，有效恢复不同程度的颜色衰减。最后，引入分块索引策略来获取更加精确的背景光估计值，并结合水下暗通道先验进一步解决水下图像的雾化模糊问题。在UIEB、RUIE等多个真实水下图像数据集上的实验结果表明，与具有代表性的水下图像增强方法相比，PSNR、SSIM指标分别提升22.17%、4.5%。Abstract: High-quality underwater optical images are crucial for tasks such as digital twins of seabed scenes, benthic habitat protection, seabed mineral resource detection, and understanding unknown underwater phenomena. However, due to factors such as complex aquatic environments and lighting conditions, underwater optical images suffer from degradation issues including color distortion, blurred details, and low contrast. Existing underwater image enhancement methods often focus on optimizing enhancement algorithms themselves, lacking systematic analytical mechanisms for tracing, classifying, and grading different types of degradation. To address this, considering the complexity and heterogeneity of underwater optical imaging environments, this paper proposes an image quality enhancement strategy that takes degradation types into account. First, a degradation-type-aware network is constructed to identify underwater hazy and blurred images, achieving an accuracy of 97%, and also demonstrating a high distinguishing capability for illumination degradation types. Second, for the identified underwater hazy images, an adaptive color correction method is designed based on the statistical distribution of color bias values in real underwater images, effectively restoring varying degrees of color attenuation. Finally, a block indexing strategy is introduced to obtain more precise background light estimates, further addressing the hazy blur issue in underwater images in conjunction with the underwater dark channel prior. Experimental results on various real underwater image datasets, including UIEB and RUIE, indicate that compared to representative underwater image enhancement methods, the PSNR and SSIM metrics are improved by 22.17% and 4.5%, respectively.
-
图 2 不同退化类型图像及其像素分布
Fig. 2 Different types of degraded images and their pixel distributions
图 6 不同程度的颜色衰减对应的色偏值范围以及
$ \alpha $ 最优补偿数值Fig. 6 The range of color decay corresponding to different degrees of color shift and the optimal compensation value for
$ \alpha $ 
图 10 水下不同退化类型图像使用同一增强算法效果
Fig. 10 Effectiveness of the same enhancement algorithm on underwater images with different degradation types
图 12 不同程度颜色退化图像的FDUM平均得分随α取值的变化情况
Fig. 12 The variation of the average FDUM score for images with different degrees of color degradation in relation to the value of α
图 13 α取值对图像主客观质量的影响示意
Fig. 13 Impact of α value on the subjective and objective quality of images
图 14 不同背景光估计方法的结果展示
Fig. 14 Results demonstration of different background light estimation methods
图 15 综合数据集上水下图像增强结果比较
Fig. 15 Comparison of underwater image enhancement results on the comprehensive dataset
图 16 UIEB数据集上水下图像增强结果比较
Fig. 16 Comparison of underwater image enhancement results on the UIEB dataset
图 17 中国近海水下图像增强结果比较
Fig. 17 Comparison of underwater image enhancement results in China’s coastal water
图 18 水下图像分割、特征点检测的应用实例结果
Fig. 18 Application examples and results of underwater image segmentation and feature point detection
表 2 UIDPNet特征图参数表
Tab. 2 UIDPNet feature map parameter table
层级 操作 输入尺寸 输出尺寸 特征图像提取 输入图像 H*W*3 224*244*8 卷积块1 ConvBlock 224*244*8 224*244*64 ConvBlock 224*244*64 − 最大池化 − 112*112*64 注意力模块 112*112*64 − 卷积块2 ConvBlock − 112*112*128 ConvBlock 112*112*128 − 最大池化 − 56*56*128 注意力模块 56*56*128 − 卷积块3 ConvBlock − 56*56*256 ConvBlock 56*56*256 − ConvBlock − − 最大池化 − 28*28*256 注意力模块 28*28*256 − 卷积块4 ConvBlock − 28*28*512 ConvBlock 28*28*512 − ConvBlock − − 最大池化 − 14*14*512 注意力模块 14*14*512 − 全局平均池化 − 1*1*512 全连接块（含dropout） 全连接层 1*1*512 1*1*256 全连接层 1*1*256 1*1*128 全连接层 1*1*128 1*1*C 
下载: 导出CSV
表 3 不同图像尺寸与迭代阈值下的耗时统计
Tab. 3 Runtime statistics for different image sizes and iteration thresholds
图像尺寸 迭代阈值 耗时（S） 400*300 500 0.334 1000 0.345 1500 0.329 640*480 500 0.530 1000 0.527 1500 0.541 960* 1200 500 1.218 1000 1.238 1500 1.206
下载: 导出CSV
表 4 无参考评价指标定量分析结果
Tab. 4 Quantitative analysis results of non-reference evaluation indicators
Methods Ancuti et al. MLLE CBLA Haze-Line TACL WaterNet 本文方法 FDUM 0.3422 0.7184 0.6214 0.6026 0.5285 0.4260 0.6674 UIQM 0.4203 1.1219 0.8873 1.2637 0.6207 0.6387 1.3410 UICQE 0.5436 0.6030 0.5328 0.5439 0.6116 0.5861 0.6074
下载: 导出CSV
表 5 全参考评价指标定量分析结果
Tab. 5 Quantitative analysis results of comprehensive evaluation indicators
Methods Ancuti et al. MLLE CBLA Haze-Line TACL WaterNet 本文方法 PSNR 15.8420 19.0382 16.2547 18.2366 18.9886 13.8584 23.2598 SSIM 0.7696 0.6891 0.7218 0.6468 0.7852 0.7688 0.8207 PCQI 0.6174 0.9847 0.7736 0.6899 0.9434 0.6820 0.9684
下载: 导出CSV
表 6 中国近海数据定量分析结果
Tab. 6 Quantitative analysis results of China’s coastal water
Methods Ancuti et al. MLLE CBLA Haze-Line TACL WaterNet 本文方法 FDUM 0.1807 0.5797 0.4126 0.3195 0.4889 0.4912 0.6755 UIQM 0.3534 3.3259 2.7129 3.3713 3.4006 2.6881 3.4955 UICQE 0.1332 0.1945 0.2316 0.1295 0.2014 0.1975 0.2288
下载: 导出CSV
-
[1] Cao Yu, Cui Xiaodong, Gan Mingyi, et al. MAL-YOLO: a lightweight algorithm for target detection in side-scan sonar images based on multi-scale feature fusion and attention mechanism[J]. International Journal of Digital Earth, 2024, 17(1): 2398050. doi: 10.1080/17538947.2024.2398050 [2] Cornwall W. Alvin, the iconic submersible, plunges deeper than ever[J]. Science, 2024, 384(6698): 833−834. [3] Cui Xiaodong, Yang Fanlin, Wu Ziyin, et al. Deep-sea sediment mixed pixel decomposition based on multibeam backscatter intensity segmentation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 4504015. doi: 10.1109/tgrs.2021.3090450 [4] Zhao Shengya, Ye Xiufen, Guo Shuxiang, et al. Deep learning-based detection of deep-sea hydrothermal polymetallic sulfides, plume, and their biological communities[C]//2024 IEEE International Conference on Mechatronics and Automation (ICMA). Tianjin, China: IEEE, 2024: 644−649. [5] 徐丹, 路航, 史金龙, 等. 基于光照先验的半监督图像增强网络在水下目标检测中的应用[J]. 海洋学报, 2025, 47(8): 69−81. doi: 10.12284/hyxb2025077Xu Dan, Lu Hang, Shi Jinlong, et al. Application of a semi-supervised image enhancement network based on lighting priors in underwater object detection[J]. Haiyang Xuebao, 2025, 47(8): 69−81. doi: 10.12284/hyxb2025077 [6] 付青青, 景春雷, 裴彦良, 等. 基于非锐化掩模引导滤波的水下图像细节增强算法研究[J]. 海洋学报, 2020, 42(7): 130−138.Fu Qingqing, Jing Chunlei, Pei Yanliang, et al. Research on underwater image detail enhancement based on unsharp mask guided filtering[J]. Haiyang Xuebao, 2020, 42(7): 130−138. [7] Shuang Xuecheng, Zhang Jin, Tian Yu. Algorithms for improving the quality of underwater optical images: a comprehensive review[J]. Signal Processing, 2024, 219: 109408. doi: 10.1016/j.sigpro.2024.109408 [8] Ancuti C O, Ancuti C, De Vleeschouwer C, et al. Color balance and fusion for underwater image enhancement[J]. IEEE Transactions on Image Processing, 2018, 27(1): 379−393. doi: 10.1109/TIP.2017.2759252 [9] Xu Shuai, Zhang Jian, Qin Xin, et al. Deep Retinex decomposition network for underwater image enhancement[J]. Computers and Electrical Engineering, 2022, 100: 107822. doi: 10.1016/j.compeleceng.2022.107822 [10] Chandrasekar A, Sreenivas M, Biswas S. PhISH-Net: physics inspired system for high resolution underwater image enhancement[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. Waikoloa, HI, USA: IEEE, 2024: 1495-1505. [11] Bi Pengfei, Xu Jian, Du Xue, et al. l2, p-norm sequential bilateral 2DPCA: a novel robust technology for underwater image classification and representation[J]. Neural Computing and Applications, 2020, 32(22): 17027−17041. doi: 10.1007/s00521-020-04936-1 [12] Qiao Weibiao, Khishe M, Ravakhah S. Underwater targets classification using local wavelet acoustic pattern and multi-layer perceptron neural network optimized by modified whale optimization algorithm[J]. Ocean Engineering, 2021, 219: 108415. doi: 10.1016/j.oceaneng.2020.108415 [13] Li Guohou, Wang Fangyuan, Zhou Ling, et al. MCANet: multi-channel attention network with multi-color space encoder for underwater image classification[J]. Computers and Electrical Engineering, 2023, 108: 108724. doi: 10.1016/j.compeleceng.2023.108724 [14] Zhang Weidong, Yu Baiqiang, Li Guohou, et al. Unified multi-color-model-learning-based deep support vector machine for underwater image classification[J]. Engineering Applications of Artificial Intelligence, 2024, 138: 109437. doi: 10.1016/j.engappai.2024.109437 [15] Chiang J Y, Chen Y C. Underwater image enhancement by wavelength compensation and dehazing[J]. IEEE Transactions on image Processing, 2012, 21(4): 1756−1769. doi: 10.1109/TIP.2011.2179666 [16] Wang Jiajie, Wan Minjie, Xu Yunkai, et al. Underwater image restoration via constrained color compensation and background light color space-based haze-line model[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 4211615. doi: 10.1109/tgrs.2024.3477911 [17] He Kaiming, Sun Jian, Tang Xiaoou. Single image haze removal using dark channel prior[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(12): 2341−2353. doi: 10.1109/TPAMI.2010.168 [18] Chao Liu, Wang Meng. Removal of water scattering[C]//2010 2nd International Conference on Computer Engineering and Technology. Chengdu: IEEE, 2010: V2-35-V2-39. [19] Drews Jr P, Do Nascimento E, Moraes F, et al. Transmission estimation in underwater single images[C]//Proceedings of the IEEE International Conference on Computer Vision Workshops. Sydney, NSW, Australia: IEEE, 2013: 825-830. [20] Li Chongyi, Guo Chunle, Ren Wenqi, et al. An underwater image enhancement benchmark dataset and beyond[J]. IEEE Transactions on Image Processing, 2020, 29: 4376−4389. doi: 10.1109/TIP.2019.2955241 [21] Liu Risheng, Fan Xin, Zhu Ming, et al. Real-world underwater enhancement: challenges, benchmarks, and solutions under natural light[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2020, 30(12): 4861−4875. doi: 10.1109/TCSVT.2019.2963772 [22] Liu Chongwei, Li Haojie, Wang Shuchang, et al. A dataset and benchmark of underwater object detection for robot picking[C]//2021 IEEE International Conference on Multimedia & Expo Workshops. Shenzhen, China: IEEE, 2021: 1−6. [23] He Dan, Cai Zhanchuan, Zhou Dujuan, et al. Inter-channel correlation modeling and improved skewed histogram shifting for reversible data hiding in color images[J]. Mathematics, 2024, 12(9): 1283. doi: 10.3390/math12091283 [24] Ulaby F T, Kouyate F, Brisco B, et al. Textural infornation in SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing, 1986, GE-24(2): 235−245. [25] Baraldi A, Panniggiani F. An investigation of the textural characteristics associated with gray level cooccurrence matrix statistical parameters[J]. IEEE Transactions on Geoscience and Remote Sensing, 1995, 33(2): 293−304. [26] Canny J. A computational approach to edge detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1986, PAMI-8(6): 679−698. [27] Buchsbaum G. A spatial processor model for object colour perception[J]. Journal of the Franklin Institute, 1980, 310(1): 1−26. doi: 10.1016/0016-0032(80)90058-7 [28] Jaffe J S. Computer modeling and the design of optimal underwater imaging systems[J]. IEEE Journal of Oceanic Engineering, 1990, 15(2): 101−111. doi: 10.1109/48.50695 [29] Tan R T. Visibility in bad weather from a single image[C]//2008 IEEE Conference on Computer Vision and Pattern Recognition. Anchorage, AK, USA: IEEE, 2008: 1−8. [30] Gibson K B, Vo D T, Nguyen T Q. An investigation of dehazing effects on image and video coding[J]. IEEE Transactions on Image Processing, 2012, 21(2): 662−673. doi: 10.1109/TIP.2011.2166968 [31] Zhang Weidong, Zhuang Peixian, Sun Haihan, et al. Underwater image enhancement via minimal color loss and locally adaptive contrast enhancement[J]. IEEE Transactions on Image Processing, 2022, 31: 3997−4010. doi: 10.1109/TIP.2022.3177129 [32] Jha M, Bhandari A K. CBLA: color-balanced locally adjustable underwater image enhancement[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 5020911. doi: 10.1109/tim.2024.3396850 [33] Berman D, Levy D, Avidan S, et al. Underwater single image color restoration using haze-lines and a new quantitative dataset[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(8): 2822−2837. doi: 10.1109/tpami.2020.2977624 [34] Liu Risheng, Jiang Zhiying, Yang Shuzhou, et al. Twin adversarial contrastive learning for underwater image enhancement and beyond[J]. IEEE Transactions on Image Processing, 2022, 31: 4922−4936. doi: 10.1109/TIP.2022.3190209 [35] Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91−110. doi: 10.1023/B:VISI.0000029664.99615.94 [36] Sheldrake T, Higgins O. Classification, segmentation and correlation of zoned minerals[J]. Computers & Geosciences, 2021, 156: 104876. doi: 10.1016/j.cageo.2021.104876 [37] Liu Yi, Jiang Qiuping, Wang Xinyi, et al. Underwater image enhancement with cascaded contrastive learning[J]. IEEE Transactions on Multimedia, 2025, 27: 1512−1525. doi: 10.1109/TMM.2024.3521739 [38] Hou Guojia, Li Nan, Zhuang Peixian, et al. Non-uniform illumination underwater image restoration via illumination channel sparsity prior[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2024, 34(2): 799−814. doi: 10.1109/TCSVT.2023.3290363 [39] Marques T P, Albu A B. L2UWE: a framework for the efficient enhancement of low-light underwater images using local contrast and multi-scale fusion[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Seattle, WA, USA: IEEE, 2020: 2286−2295. -
计量
- 文章访问数: 18
- HTML全文浏览量: 13
- PDF下载量: 0
- 被引次数: 0
