吉林大学学报(工学版) ›› 2020, Vol. 50 ›› Issue (5): 1771-1777.doi: 10.13229/j.cnki.jdxbgxb20190443

• 计算机科学与技术 • 上一篇    

基于卷积神经网络的联合估计图像去雾算法

王柯俨(),王迪,赵熹,陈静怡,李云松   

  1. 西安电子科技大学 综合业务网理论及关键技术国家重点实验室,西安 710071
  • 收稿日期:2019-05-09 出版日期:2020-09-01 发布日期:2020-09-16
  • 作者简介:王柯俨(1980-),女,副教授,博士.研究方向:图像编码与处理.E-mail:kywang@mail.xidian.edu.cn
  • 基金资助:
    国家自然科学基金项目(61301291);高等学校学科创新引智计划(“111”计划”)项目(B08038)

Image dehazing based on joint estimation via convolutional neural network

Ke-yan WANG(),Di WANG,Xi ZHAO,Jing-yi CHEN,Yun-song LI   

  1. State Key Laboratory of Integrated Service Networks, Xidian University, Xi'an 710071, China
  • Received:2019-05-09 Online:2020-09-01 Published:2020-09-16

摘要:

室外拍摄的图像由于空气中的大气颗粒会具有较低的对比度和能见度,影响主观视觉和图像处理系统的有效性,为此本文提出了一种端到端的透射率和大气光联合估计去雾网络。通过共享特征模块获取透射率和大气光共有的全局特征,利用金字塔池化模块的多尺度卷积提取组合特征;然后,通过两个并行的分支分别估计透射率和大气光;最后,通过大气散射模型反演出无雾图像。实验结果表明:本文方法恢复图像较其他去雾方法的对比度更强,色彩更自然,网络优化参数更少。

关键词: 信息处理技术, 图像去雾, 卷积神经网络, 联合估计, 大气散射模型

Abstract:

Images captured in outdoor scenes typically have lower contrast and visibility due to atmospheric particles suspended in the air, directly affecting the effectiveness of subjective vision and intelligent image processing systems. To solve this problem, we propose an end-to-end network, which jointly estimates transmission map and atmospheric light effectively. The global characteristics of transmittance and atmospheric light are acquired simultaneously by shared feature module, and the multi-scale convolution of the pyramid pooling module is utilized to extract the combination feature. Then, the transmittance and atmospheric light are separately estimated by two parallel branch networks. Finally, the fog-free image is obtained by inversion of the atmospheric scattering model. The experiment results show that proposed method removes haze more effectively compared to these state-of-the-art methods, and the restored image has higher contrast and more natural color. Moreover, the proposed network has less parameters.

Key words: information processing technology, image dehazing, convolutional neural network, joint estimation, atmospheric scattering model

中图分类号: 

  • TP751.1

图1

大气散射模型示意图"

图2

透射率及大气光值联合估计算法流程"

图3

网络结构图"

图4

金字塔池化模块结构图"

图5

基于深度学习的去雾算法在Reside合成雾图上的去雾效果对比"

表1

不同类型合成雾图下去雾结果对比"

DatasetMetricsDehaze NetMSCNNBilinear -NetAOD-NetOurs
Make3DPSNR20.0122.6416.2119.3723.15
SSIM0.870.940.590.870.94
NYUPSNR19.0319.0514.1516.3419.05
SSIM0.860.660.860.800.87

图6

基于深度学习的去雾算法在真实雾图上的效果对比"

1 Wei Y, Yuan Q, Shen H, et al. A universal remote sensing image quality improvement method with deep learning[C]∥IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Beijing, 2016: 6950-6953.
2 Zhang Q, Yuan Q, Zeng C, et al. Missing data reconstruction in remote sensing image with a unified spatial-temporal-spectral deep convolutional neural network[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(8): 4274-4288.
3 Enomoto K, Sakurada K, Wang W, et al. Filmy cloud removal on satellite imagery with multispectral conditional generative adversarial nets[C]∥Conference on Computer Vision and Pattern Recognition Workshops. Honolulu: HI, 2017: 1533-1541.
4 Kim T K, Paik J K, Kang B S. Contrast enhancement system using spatially adaptive histogram equalization with temporal filtering[J]. IEEE Transactions on Consumer Electronics, 1998, 44(1): 82-87.
5 Stark J A. Adaptive image contrast enhancement using generalizations of histogram equalization[J]. IEEE Transactions on Image Processing A Publication of the IEEE Signal Processing Society, 2000, 9(5): 889-896.
6 He K, Sun J, Tang X. Single image haze removal using dark channel prior[J]. IEEE Transactionson on Pattern Analysis and Machine Intelligence, 2011, 33(12): 2341-2353.
7 谭华春, 朱湧, 赵亚男, 等. 图像去雾的大气光幕修补改进算法[J]. 吉林大学学报: 工学版, 2013, 43(): 389-393.
Tan Hua-chun, Zhu Yong, Zhao Ya-nan, et al. Image fog removal using improved atmospheric veil inpainting[J]. Journal of Jilin University(Engineering and Technology Edition), 2013, 43(Sup.1): 389-393.
8 王柯俨, 胡妍, 王怀, 等. 结合天空分割和超像素级暗通道的图像去雾算法[J]. 吉林大学学报: 工学版, 2019, 49(4): 1377-1384.
Wang Ke⁃yan, Hu Yan, Wang Huai, et al. Image dehazing algorithm by sky segmentation and superpixel⁃level dark channel[J]. Journal of Jilin University(Engineering and Technology Edition), 2019, 49(4): 1377-1384.
9 Cai B, Xu X, Jia K, et al. Dehazenet: an end-to-end system for single image haze removal[J]. IEEE Transactions on Image Processing, 2016, 25(11): 5187-5198.
10 Ren W, Liu S, Zhang H, et al. Single image dehazing via multi-scale convolutional neural networks[C]∥European Conference on Computer Vision, Springer, Cham, 2016: 154-169.
11 Yang H, Pan J, Yan Q, et al. Image dehazing using bilinear composition loss function[EB/OL]. [2019-04-30].
12 Li B, Peng X, Wang Z, et al. Aod-net: all-in-one dehazing network[C]∥IEEE International Conference on Computer Vision, Venice, 2017: 4770-4778.
13 McCartney E J. Optics of the atmosphere: scattering by molecules and particles[J]. New York: Wiley, 1976, 76: 23-32.
14 Deng J, Dong W, Socher R, et al. ImageNet: a large-scale hierarchical image database[C]∥IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Miami, FL, 2009: 248-255.
15 Everingham M, Gool L V, Williams C K I, et al. The pascal visual object classes(VOC) challenge[J]. International Journal of Computer Vision, 2010, 88(2): 303-338.
16 Liu F, Shen C, Lin G. Deep convolutional neural fields for depth estimation from a single image[C]∥IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, 2015: 5162-5170.
17 Kingma D P, Ba J. Adam: a method for stochastic optimization[EB/OL]. [2019-04-30].
18 Li B, Ren W, Fu D, et al. Benchmarking single image dehazing and beyond[J]. IEEE Transactions on Image Processing, 2019, 28(1): 492-505.
19 Wang Z, Bovik A C, Sheikh H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004, 13(4): 600-612.
20 Saxena A, Sun M, Ng A Y. Make3d: learning 3d scene structure from a single still image[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, 31(5): 824-840.
21 Silberman N, Hoiem D, Kohli P, et al. Indoor segmentation and support inference from RGBD images[C]∥European Conference on Computer Vision, Berlin, 2012: 746-760.
[1] 史再峰,李金卓,曹清洁,李慧龙,胡起星. 基于生成对抗网络的低剂量能谱层析成像去噪算法[J]. 吉林大学学报(工学版), 2020, 50(5): 1755-1764.
[2] 车翔玖,董有政. 基于多尺度信息融合的图像识别改进算法[J]. 吉林大学学报(工学版), 2020, 50(5): 1747-1754.
[3] 张薇,韩勇,金铭,乔晓林. 基于托普利兹矩阵集重构的相干信源波达方向估计[J]. 吉林大学学报(工学版), 2020, 50(2): 703-710.
[4] 程艳芬,姚丽娟,袁巧,陈先桥. 水下视频图像清晰化方法[J]. 吉林大学学报(工学版), 2020, 50(2): 668-677.
[5] 于晓辉,张志成,李新波,孙晓东. 基于状态空间模型的指数衰减正弦信号参数估计[J]. 吉林大学学报(工学版), 2019, 49(6): 2083-2088.
[6] 车翔玖,刘华罗,邵庆彬. 基于Fast RCNN改进的布匹瑕疵识别算法[J]. 吉林大学学报(工学版), 2019, 49(6): 2038-2044.
[7] 周柚,杨森,李大琳,吴春国,王岩,王康平. 基于现场可编程门电路的人脸检测识别加速平台[J]. 吉林大学学报(工学版), 2019, 49(6): 2051-2057.
[8] 刘富, 权美静, 王柯, 刘云, 康冰, 韩志武, 侯涛. 仿蝎子振源定位机理的位置指纹室内定位方法[J]. 吉林大学学报(工学版), 2019, 49(6): 2076-2082.
[9] 赵宏伟,王鹏,范丽丽,胡黄水,刘萍萍. 相似性保持实例检索方法[J]. 吉林大学学报(工学版), 2019, 49(6): 2045-2050.
[10] 马子骥,卢浩,董艳茹. 双通道单图像超分辨率卷积神经网络[J]. 吉林大学学报(工学版), 2019, 49(6): 2089-2097.
[11] 郭继昌,吴洁,郭春乐,朱明辉. 基于残差连接卷积神经网络的图像超分辨率重构[J]. 吉林大学学报(工学版), 2019, 49(5): 1726-1734.
[12] 曹运合,曾丽,王宇. 基于特征空间的子阵级自适应和差波束测角方法[J]. 吉林大学学报(工学版), 2019, 49(5): 1735-1744.
[13] 卢洋,王世刚,赵文婷,赵岩. 基于离散Shearlet类别可分性测度的人脸表情识别方法[J]. 吉林大学学报(工学版), 2019, 49(5): 1715-1725.
[14] 董超,刘晶红,徐芳,王仁浩. 光学遥感图像舰船目标快速检测方法[J]. 吉林大学学报(工学版), 2019, 49(4): 1369-1376.
[15] 王柯俨,胡妍,王怀,李云松. 结合天空分割和超像素级暗通道的图像去雾算法[J]. 吉林大学学报(工学版), 2019, 49(4): 1377-1384.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!