高级搜索

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于多先验约束和一致性正则的半监督图像去雾算法

苏延召 何川 崔智高 姜柯 蔡艳平 李艾华

苏延召, 何川, 崔智高, 姜柯, 蔡艳平, 李艾华. 基于多先验约束和一致性正则的半监督图像去雾算法[J]. 电子与信息学报, 2022, 44(10): 3409-3418. doi: 10.11999/JEIT220381
引用本文: 苏延召, 何川, 崔智高, 姜柯, 蔡艳平, 李艾华. 基于多先验约束和一致性正则的半监督图像去雾算法[J]. 电子与信息学报, 2022, 44(10): 3409-3418. doi: 10.11999/JEIT220381
SU Yanzhao, HE Chuan, CUI Zhigao, JIANG Ke, CAI Yanping, LI Aihua. Semi-supervised Image Dehazing Algorithm Based on Multi-prior Constraint and Consistency Regularization[J]. Journal of Electronics & Information Technology, 2022, 44(10): 3409-3418. doi: 10.11999/JEIT220381
Citation: SU Yanzhao, HE Chuan, CUI Zhigao, JIANG Ke, CAI Yanping, LI Aihua. Semi-supervised Image Dehazing Algorithm Based on Multi-prior Constraint and Consistency Regularization[J]. Journal of Electronics & Information Technology, 2022, 44(10): 3409-3418. doi: 10.11999/JEIT220381

基于多先验约束和一致性正则的半监督图像去雾算法

doi: 10.11999/JEIT220381
基金项目: 国家自然科学基金 (61773389),中国博士后基金(2019M663635),陕西省青年科技之星计划(2021KJXX-22),陕西省自然科学基金(2020JQ-2)
详细信息
    作者简介:

    苏延召:男,讲师,研究方向为计算机视觉、图像处理

    何川:男,副教授,研究方向为计算机视觉、图像处理

    崔智高:男,副教授,研究方向为计算机视觉、图像处理

    姜柯:男,讲师,研究方向为计算机视觉、图像处理

    蔡艳平:男,教授,研究方向为图像处理与机器视觉

    李艾华:男,教授,研究方向为图像处理与机器视觉

    通讯作者:

    崔智高 cuizg10@126.com

  • 中图分类号: TN911.73; TP391

Semi-supervised Image Dehazing Algorithm Based on Multi-prior Constraint and Consistency Regularization

Funds: The National Natural Science Foundation of China (61773389), The Postdoctoral Science Foundation of China (2019M663635), The Shaanxi Young Science and Technology Star (2021KJXX-22), The Natural Science Foundation of Shaanxi Province (2020JQ-2)
  • 摘要: 针对合成雾霾图像训练的去雾模型在真实场景中去雾效果不佳、对高层视觉任务性能提升不明显等问题,该文提出一种基于多先验约束和一致性正则的半监督图像去雾算法。该方法采用编码器-解码器网络结构,同时在合成雾霾图像与真实雾霾图像上学习去雾映射,并利用多种统计先验去雾结果作为真实雾霾图像参考真值进行半监督学习,同时通过多张真实雾霾图像的随机混合进行一致性正则约束,以消除多种先验去雾结果差异以及噪声干扰,提高图像去雾结果的视觉质量。实验对比结果表明,所提算法可比现有方法获得更好的真实场景去雾结果,并且能够显著提升高层视觉任务性能。
  • 图  1  本文算法总体框架示意图

    图  2  不同先验去雾方法处理结果对比

    图  3  本文方法与现有代表性方法在合成数据集上的去雾处理结果对比

    图  4  本文方法与现有代表性方法在真实场景中的去雾处理结果对比

    图  5  本文方法在其他场景去雾霾结果

    表  1  本文图像去雾网络结构详细设计表(数字表示图像序号)

    Conv1Conv2Conv3Conv4Res1-Res15UpConv1UpConv2UpConv3Conv5
    输入3326412825625625612864
    输出326412825625612864323
    卷积核大小744433333
    卷积步长122211111
    边界填充311111111
    下载: 导出CSV

    表  2  图像去雾定量实验结果对比(红色表示第1,绿色表示第2,蓝色表示第3)

    方法数据集
    SOTS(indoor)SOTS (outdoor)HAZERDIHAZEOHAZEBeDDE
    PSNR (dB)SSIMPSNR (dB)SSIMPSNR (dB)SSIMPSNR (dB)SSIMPSNR (dB)SSIMVIRI
    BDCP14.8710.742517.4590.818613.0180.780215.8930.8064 15.4750.72880.89700.9650
    NLD17.3280.80518.1150.87014.5710.800112.6300.62816.0800.7220.85920.9583
    MSBDN31.5690.982530.2550.963014.7580.795616.5470.806218.3530.69990.84010.9655
    SED22.2340.920025.2950.942016.1320.838915.8540.754218.5220.74140.88140.9674
    DAAD25.6850.952925.380.910917.0160.816517.3690.825618.8870.77810.88350.9652
    PSD12.4960.717715.5780.804914.2120.771212.5450.736412.5130.70900.83920.9640
    本文方法25.2390.958424.8410.939416.4590.845616.8180.813017.6030.79990.89720.9659
    下载: 导出CSV

    表  3  雾霾图像目标检测实验结果对比(红色表示第1,绿色表示第2,蓝色表示第3)

    方法目标类别
    PersonbicyclecarmotorbikebusAllGain
    APAPAPAPAPmAP
    Hazy Image81.7665.7675.0663.1337.5964.46
    BDCP82.2363.9272.9958.4143.3864.19–0.27
    NLD80.4863.4473.8759.1935.9362.58–1.88
    MSBDN83.0065.4775.6661.5538.4864.83+0.27
    SED82.4365.5975.8161.9238.9964.94+0.48
    DAAD81.4564.2776.3461.8740.7464.93+0.47
    PSD82.7865.8274.8160.1642.1865.15+0.69
    本文方法82.8965.6975.6662.3742.0965.74+1.28
    下载: 导出CSV

    表  4  消融实验结果对比

    数据集指标名称基准方法变体1变体2变体3变体4本文方法
    BeDDEVI0.87880.88120.88540.89030.89350.8972
    RI0.95520.95890.96030.96150.96320.9659
    RTTSmAP (%)64.764.663.965.265.465.7
    下载: 导出CSV

    表  5  5种去雾方法运行时间对比(s)

    MSBDNSEDDAADPSD本文方法
    运行时间0.0750.0120.0490.0430.028
    下载: 导出CSV
  • [1] 李红云, 施云, 高银. 基于显著性权重的多曝光融合的单幅雾天图像复原算法[J]. 电子与信息学报, 2022, 44(1): 261–270. doi: 10.11999/JEIT200931

    LI Hongyun, SHI Yun, and GAO Yin. Single image dehazing via saliency weighted multi-exposure fusion[J]. Journal of Electronics &Information Technology, 2022, 44(1): 261–270. doi: 10.11999/JEIT200931
    [2] YANG Wenhan, YUAN Ye, and REN Wenqi, et al. Advancing image understanding in poor visibility environments: A collective benchmark study[J]. IEEE Transactions on Image Processing, 2020, 29: 5737–5752. doi: 10.1109/TIP.2020.2981922
    [3] NAYAR S K and NARASIMHAN S G. Vision in bad weather[C]. Proceedings of the Seventh IEEE International Conference on Computer Vision, Kerkyra, Greece, 1999: 820–827.
    [4] HE Kaiming, SUN Jian, and 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
    [5] ZHU Qingsong, MAI Jiaming, and SHAO Ling. A fast single image haze removal algorithm using color attenuation prior[J]. IEEE Transactions on Image Processing, 2015, 24(11): 3522–3533. doi: 10.1109/TIP.2015.2446191
    [6] FATTAL R. Dehazing using color-lines[J]. ACM Transactions on Graphics, 2014, 34(1): 13. doi: 10.1145/2651362
    [7] REN Wenqi, PAN Jinshan, ZHANG Hua, et al. Single image dehazing via multi-scale convolutional neural networks with holistic edges[J]. International Journal of Computer Vision, 2020, 128(1): 240–259. doi: 10.1007/s11263-019-01235-8
    [8] CAI Bolun, XU Xiangmin, JIA Kui, et al. DehazeNet: An end-to-end system for single image haze removal[J]. IEEE Transactions on Image Processing, 2016, 25(11): 5187–5198. doi: 10.1109/TIP.2016.2598681
    [9] ZHANG He and PATEL V M. Densely connected pyramid dehazing network[C]. The 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, 2018: 3194–3203.
    [10] LIU Xiaohong, MA Yongrui, SHI Zhihao, et al. GridDehazeNet: Attention-based multi-scale network for image dehazing[C]. IEEE/CVF International Conference on Computer Vision, Seoul, Korea (South), 2019: 7313–7322.
    [11] QIN Xu, WANG Zhilin, BAI Yuanchao, et al. FFA-Net: Feature fusion attention network for single image dehazing[C]. 34th AAAI Conference on Artificial Intelligence, New York, USA, 2020: 11908–11915.
    [12] DONG Hang, PAN Jinshan, XIANG Lei, et al. Multi-scale boosted dehazing network with dense feature fusion[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, USA, 2020: 2154–2164.
    [13] SU Yanzhao, CUI Zhigao, HE Chuan, et al. Prior guided conditional generative adversarial network for single image dehazing[J]. Neurocomputing, 2021, 423: 620–638. doi: 10.1016/j.neucom.2020.10.061
    [14] LI Lerenhan, DONG Yunlong, REN Wenqi, et al. Semi-supervised image dehazing[J]. IEEE Transactions on Image Processing, 2020, 29: 2766–2779. doi: 10.1109/TIP.2019.2952690
    [15] SHAO Yuanjie, LI Lerenhan, REN Wenqi, et al. Domain adaptation for image dehazing[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, USA, 2020: 2805–2814.
    [16] CHEN Zeyuan, WANG Yangchao, YANG Yang, et al. PSD: Principled synthetic-to-real dehazing guided by physical priors[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition, Nashville, USA, 2021: 7176–7185.
    [17] 张世辉, 路佳琪, 宋丹丹, 等. 基于多尺度特征结合细节恢复的单幅图像去雾方法[J]. 电子与信息学报, 待发表.

    ZHANG Shihui, LU Jiaqi, SONG Dandan, et al. Single image dehazing method based on multi-scale features combined with detail recovery[J]. Journal of Electronics & Information Technology. To be published,
    [18] WANG Nian, CUI Zhigao, SU Yanzhao, et al. Multiscale supervision-guided context aggregation network for single image dehazing[J]. IEEE Signal Processing Letters, 2022, 29: 70–74. doi: 10.1109/LSP.2021.3125272
    [19] MEI Kangfu, JIANG Aiwen, LI Juncheng, et al. Progressive feature fusion network for realistic image dehazing[C]. 14th Asian Conference on Computer Vision, Perth, Australia, 2019: 203–215. DOI: 10.1007/978-3-030-20887-5_13.
    [20] ZHAO Hang, GALLO O, FROSIO I, et al. Loss functions for image restoration with neural networks[J]. IEEE Transactions on Computational Imaging, 2017, 3(1): 47–57. doi: 10.1109/TCI.2016.2644865
    [21] 卢辉斌, 赵燕芳, 赵永杰, 等. 基于亮通道和暗通道结合的图像去雾[J]. 光学学报, 2018, 38(11): 1115004. doi: 10.3788/AOS201838.111500

    LU Huibin, ZHAO Yanfang, ZHAO Yongjie, et al. Image defogging based on combination of image bright and dark channels[J]. Acta Optica Sinica, 2018, 38(11): 1115004. doi: 10.3788/AOS201838.111500
    [22] BERMAN D, TREIBITZ T, and AVIDAN S. Non-local image dehazing[C]. The 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, 2016: 1674–1682.
    [23] GHIASI G, ZOPH B, CUBUK E D, et al. Multi-task self-training for learning general representations[C]. IEEE/CVF International Conference on Computer Vision, Montreal, Canada, 2021: 8836–8845.
    [24] YUN S, HAN D, CHUN S, et al. CutMix: Regularization strategy to train strong classifiers with localizable features[C]. IEEE/CVF International Conference on Computer Vision, Seoul, Korea (South), 2019: 6022–6031.
    [25] REN Shaoqing, HE Kaiming, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137–1149. doi: 10.1109/tpami.2016.2577031
    [26] LI Boyi, REN Wenqi, FU Dengpan, et al. Benchmarking single-image dehazing and beyond[J]. IEEE Transactions on Image Processing, 2019, 28(1): 492–505. doi: 10.1109/TIP.2018.2867951
    [27] ZHANG Yanfu, DING Li, and SHARMA G. HazeRD: An outdoor scene dataset and benchmark for single image dehazing[C]. 2017 IEEE International Conference on Image Processing (ICIP), Beijing, China, 2017: 3205–3209.
    [28] ANCUTI C, ANCUTI C O, TIMOFTE R, et al. I-HAZE: A dehazing benchmark with real hazy and haze-free indoor images[C]. The 19th International Conference on Advanced Concepts for Intelligent Vision Systems, Poitiers, France, 2018: 620–631.
    [29] ANCUTI C O, ANCUTI C, TIMOFTE R, et al. O-HAZE: A dehazing benchmark with real hazy and haze-free outdoor images[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Salt Lake City, USA, 2018: 867–875.
    [30] ZHAO Shiyu, ZHANG Lin, HUANG Shuaiyi, et al. Dehazing evaluation: Real-world benchmark datasets, criteria, and baselines[J]. IEEE Transactions on Image Processing, 2020, 29: 6947–6962. doi: 10.1109/TIP.2020.2995264
    [31] REDMON J and FARHADI A. YOLOv3: An incremental improvement[EB/OL].https://doi.org/10.48550/arXiv.1804.02767, 2018.
    [32] 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
  • 加载中
图(5) / 表(5)
计量
  • 文章访问数:  583
  • HTML全文浏览量:  291
  • PDF下载量:  115
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-04-01
  • 修回日期:  2022-08-25
  • 网络出版日期:  2022-09-14
  • 刊出日期:  2022-10-19

目录

    /

    返回文章
    返回