《计算机应用》唯一官方网站 ›› 2021, Vol. 41 ›› Issue (11): 3345-3352.DOI: 10.11772/j.issn.1001-9081.2020121898
收稿日期:
2020-12-04
修回日期:
2021-05-13
接受日期:
2021-08-03
发布日期:
2021-05-13
出版日期:
2021-11-10
通讯作者:
蒋慕蓉
作者简介:
李福海(1994—),男,重庆人,硕士研究生,主要研究方向:深度学习、图像重建基金资助:
Fuhai LI1, Murong JIANG1(), Lei YANG2, Junyi CHEN2
Received:
2020-12-04
Revised:
2021-05-13
Accepted:
2021-08-03
Online:
2021-05-13
Published:
2021-11-10
Contact:
Murong JIANG
About author:
LI Fuhai,born in 1994,M. S. candidate. His research interests
include deep learning,image reconstructionSupported by:
摘要:
针对云南天文台拍摄的高度模糊的太阳斑点图像采用现有深度学习算法恢复难度大、高频信息难以重建等问题,提出了一种基于生成对抗网络(GAN)与梯度信息联合的去模糊方法来重建太阳斑点图,并很好地恢复出图像的高频信息。该方法由一个生成器与两个鉴别器构成:首先,生成器采用特征金字塔网络(FPN)框架来获取图像多尺度特征,再将这些特征分层次输入梯度分支以梯度图的形式捕获更小的局部特征;然后,联合梯度分支结果与FPN结果共同重建出具有高频信息的太阳斑点图像;其次,在常规对抗鉴别器的基础上,增加了一个鉴别器用于保证由梯度分支产生的梯度图更加真实;最后,引入一个包括像素内容损失、感知损失和对抗损失的联合训练损失来引导模型进行太阳斑点图像高分辨率重建。实验结果表明,进行图像预处理后的所提方法与现有的深度学习去模糊方法相比,高频信息恢复能力更强,峰值信噪比(PSNR)和结构相似性(SSIM)指标均有显著提高,分别达到27.801 0 dB与0.851 0,能够满足太阳观测图像高分辨率重建的需要。
中图分类号:
李福海, 蒋慕蓉, 杨磊, 谌俊毅. 基于生成对抗网络的梯度引导太阳斑点图像去模糊方法[J]. 计算机应用, 2021, 41(11): 3345-3352.
Fuhai LI, Murong JIANG, Lei YANG, Junyi CHEN. Solar speckle image deblurring method with gradient guidance based on generative adversarial network[J]. Journal of Computer Applications, 2021, 41(11): 3345-3352.
方法 | PSNR/dB | SSIM |
---|---|---|
SPSR | 22.965 5 | 0.638 3 |
DRN-L | 24.274 2 | 0.652 4 |
DeblurGANv2-Mobile | 24.476 4 | 0.652 6 |
DeblurGANv2-Inception | 24.064 0 | 0.674 1 |
Cycle-GAN1 | 25.123 7 | 0.744 2 |
Cycle-GAN2 | 23.620 1 | 0.694 7 |
本文方法 | 27.603 9 | 0.833 4 |
本文方法+预处理 | 27.801 0 | 0.851 0 |
表1 不同方法在测试集上的评估结果
Tab. 1 Evaluation results of different methods on test set
方法 | PSNR/dB | SSIM |
---|---|---|
SPSR | 22.965 5 | 0.638 3 |
DRN-L | 24.274 2 | 0.652 4 |
DeblurGANv2-Mobile | 24.476 4 | 0.652 6 |
DeblurGANv2-Inception | 24.064 0 | 0.674 1 |
Cycle-GAN1 | 25.123 7 | 0.744 2 |
Cycle-GAN2 | 23.620 1 | 0.694 7 |
本文方法 | 27.603 9 | 0.833 4 |
本文方法+预处理 | 27.801 0 | 0.851 0 |
方法 | SSIM | 模型尺寸/MB | 训练时间/h |
---|---|---|---|
无预处理 | 0.833 4 | 271.3 | 86.8 |
Mobile | 0.851 0 | 291.5 | 103.8 |
Inception | 0.853 2 | 510.7 | 143.2 |
表2 不同预处理方法的性能比较
Tab. 2 Performance comparison of different preprocessing methods
方法 | SSIM | 模型尺寸/MB | 训练时间/h |
---|---|---|---|
无预处理 | 0.833 4 | 271.3 | 86.8 |
Mobile | 0.851 0 | 291.5 | 103.8 |
Inception | 0.853 2 | 510.7 | 143.2 |
方法 | PSNR/dB | SSIM |
---|---|---|
FPN编码器分支 | 25.261 6 | 0.775 1 |
去掉分支 | 24.840 6 | 0.727 5 |
表3 不同梯度分支构造带来的影响
Tab. 3 Impact of different gradient branch structures
方法 | PSNR/dB | SSIM |
---|---|---|
FPN编码器分支 | 25.261 6 | 0.775 1 |
去掉分支 | 24.840 6 | 0.727 5 |
方法 | PSNR/dB | SSIM |
---|---|---|
DeblurGANv2-Mobile | 28.54 | 0.929 4 |
DeblurGANv2-Inception | 28.85 | 0.932 7 |
本文方法 | 28.69 | 0.931 2 |
本文方法+预处理 | 28.76 | 0.921 8 |
表4 不同方法在DVD数据集上的评估结果
Tab. 4 Evaluation results of different methods on DVD dataset
方法 | PSNR/dB | SSIM |
---|---|---|
DeblurGANv2-Mobile | 28.54 | 0.929 4 |
DeblurGANv2-Inception | 28.85 | 0.932 7 |
本文方法 | 28.69 | 0.931 2 |
本文方法+预处理 | 28.76 | 0.921 8 |
方法 | PSNR/dB | SSIM |
---|---|---|
DL | 24.64 | 0.841 9 |
DeepDeblur | 29.08 | 0.913 5 |
SRN | 30.26 | 0.934 2 |
DeblurGANv2-Mobile | 28.17 | 0.925 4 |
DeblurGANv2-Inception | 29.55 | 0.934 4 |
本文方法 | 28.85 | 0.921 2 |
本文方法+预处理 | 28.92 | 0.923 2 |
表5 不同方法在GOPRO数据集上的评估结果
Tab. 5 Evaluation results of different methods on GOPRO dataset
方法 | PSNR/dB | SSIM |
---|---|---|
DL | 24.64 | 0.841 9 |
DeepDeblur | 29.08 | 0.913 5 |
SRN | 30.26 | 0.934 2 |
DeblurGANv2-Mobile | 28.17 | 0.925 4 |
DeblurGANv2-Inception | 29.55 | 0.934 4 |
本文方法 | 28.85 | 0.921 2 |
本文方法+预处理 | 28.92 | 0.923 2 |
1 | 张兰强.太阳高分辨率力成像多层共轭自适应光学技术研究[D].北京:中国科学院大学,2014:20-28. |
ZHANG L Q. Research on multi-conjugate adaptive optics technology for solar high-resolution force imaging [D]. Beijing: University of Chinese Academy of Sciences, 2014: 20-28. | |
2 | 刘忠.天文图象高分辨重建及空域性质研究[D].北京:中国科学院大学,2003:25-39. |
LIU Z. Research on high-resolution image reconstruction and spatial characteristics of astronomical images [D]. Beijing: University of Chinese Academy of Sciences, 2003: 25-39. | |
3 | 邱耀辉,刘忠,卢汝为,等.天文图像空域重建新方法:迭代位移叠加法[J].光学学报,2001,21(2):186-191. 10.3321/j.issn:0253-2239.2001.02.015 |
QIU Y H, LIU Z, LU R W, et al. A new method for astronomical image reconstruction in spatial region: iterative shift-and-add technique [J]. Acta Optica Sinica, 2001, 21(2): 186-191. 10.3321/j.issn:0253-2239.2001.02.015 | |
4 | 向永源.太阳高分辨高速重建算法的研究[D].北京:中国科学院大学,2016:19-22. |
XIANG Y Y. Research on high speed high resolution solar reconstruction algorithm [D]. Beijing: University of Chinese Academy of Sciences, 2016: 19-22. | |
5 | LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection [C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 936-944. 10.1109/cvpr.2017.106 |
6 | MA C, RAO Y M, CHENG Y, et al. Structure-preserving super resolution with gradient guidance [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 7766-7775. 10.1109/cvpr42600.2020.00779 |
7 | JOHNSON J, ALAHI A, LI F F. Perceptual losses for real-time style transfer and super-resolution [C]// Proceedings of the 2016 European Conference on Computer Vision, LNCS9906. Cham: Springer, 2016: 1604-1613. |
8 | GU J J, LU H N, ZUO W M, et al. Blind super-resolution with iterative kernel correction [C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 1604-1613. 10.1109/cvpr.2019.00170 |
9 | GUO Y, CHEN J, WANG J D, et al. Closed-loop matters: dual regression networks for single image super-resolution [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 5406-5415. 10.1109/cvpr42600.2020.00545 |
10 | KAUFMAN A, FATTAL R. Deblurring using analysis synthesis networks pair [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 5810-5819. 10.1109/cvpr42600.2020.00585 |
11 | YANG F Z, YANG H, FU J L, et al. Learning texture transformer network for image super-resolution [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 5790-5799. 10.1109/cvpr42600.2020.00583 |
12 | GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial networks [C]// Proceedings of the 27th International Conference on Neural Information Processing Systems. Cambridge: MIT Press, 2014: 2672-2680. |
13 | RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks [EB/OL]. (2016-01-07) [2020-08-27]. . |
14 | ZHU J Y, PARK T, ISOLA P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks [C]// Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 2242-2251. 10.1109/iccv.2017.244 |
15 | ENGIN D, GENC A, EKENEL H K. Cycle-dehaze: enhanced CycleGAN for single image dehazing [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Piscataway: IEEE, 2018: 938-946. 10.1109/cvprw.2018.00127 |
16 | KUPYN O, MARTYNIUK T, WU J R, et al. DeblurGAN-v2: deblurring (orders-of-magnitude) faster and better [C]// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 8877-8886. 10.1109/iccv.2019.00897 |
17 | JOLICOEUR-MARTINEAU A. The relativistic discriminator: a key element missing from standard GAN [EB/OL]. (2018-09-10) [2020-08-27]. . |
18 | GULRAJANI I, AHMED F, ARJOVSKY M, et al. Improved training of Wasserstein GANs [C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2017: 5769-5779. |
19 | REN Y H, JIANG M R, FU Z X, et al. Reconstruction of single-frame solar speckle image with cycle consistency loss and perceptual loss [C]// Proceedings of the 2019 6th International Conference on Information Science and Control Engineering. Piscataway: IEEE, 2019: 439-443. 10.1109/icisce48695.2019.00094 |
20 | JIA P, HUANG Y, CAI B J, et al. Solar image restoration with the CycleGAN based on multi-fractal properties of texture features [J]. The Astrophysical Journal Letters, 2019, 881(2): Article No.L30. 10.3847/2041-8213/ab365f |
21 | WANG X T, YU K, WU S X, et al. ESRGAN: enhanced super resolution generative adversarial networks [C]// Proceedings of the 2018 European Conference on Computer Vision, LNCS11133. Cham: Springer, 2018: 63-79. |
22 | SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition [EB/OL]. (2015-04-10) [2020-08-27]. . 10.5244/c.28.6 |
23 | LEDIG C, THEIS L, HUSZÁR F, et al. Photo-realistic single image super-resolution using a generative adversarial network [C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 105-114. 10.1109/cvpr.2017.19 |
24 | SUN J, CAO W F, XU Z B, et al. Learning a convolutional neural network for non-uniform motion blur removal [C]// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2015: 769-777. 10.1109/cvpr.2015.7298677 |
25 | NAH S, KIM T H, LEE K M. Deep multi-scale convolutional neural network for dynamic scene deblurring [C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 257-265. 10.1109/cvpr.2017.35 |
26 | TAO X, GAO H Y, SHEN X Y, et al. Scale-recurrent network for deep image deblurring [C]// Proceedings of the 2018 IEEE/CVG Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 8174-8182. 10.1109/cvpr.2018.00853 |
[1] | 刘丽, 侯海金, 王安红, 张涛. 基于多尺度注意力的生成式信息隐藏算法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2102-2109. |
[2] | 王昊冉, 于丹, 杨玉丽, 马垚, 陈永乐. 面向工控系统未知攻击的域迁移入侵检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1158-1165. |
[3] | 郑毅, 廖存燚, 张天倩, 王骥, 刘守印. 面向城区的基于图去噪的小区级RSRP估计方法[J]. 《计算机应用》唯一官方网站, 2024, 44(3): 855-862. |
[4] | 周辉, 陈玉玲, 王学伟, 张洋文, 何建江. 基于生成对抗网络的联邦学习深度影子防御方案[J]. 《计算机应用》唯一官方网站, 2024, 44(1): 223-232. |
[5] | 陈少权, 蔡剑平, 孙岚. 动态梯度阈值裁剪的差分隐私生成对抗网络算法[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2065-2072. |
[6] | 刘安阳, 赵怀慈, 蔡文龙, 许泽超, 解瑞灯. 基于主动判别机制的自适应生成对抗网络图像去模糊算法[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2288-2294. |
[7] | 靳鑫, 刘仰川, 朱叶晨, 张子健, 高欣. 基于残差编解码-生成对抗网络的正弦图修复的稀疏角度锥束CT图像重建[J]. 《计算机应用》唯一官方网站, 2023, 43(6): 1950-1957. |
[8] | 吴家皋, 章仕稳, 蒋宇栋, 刘林峰. 基于状态精细化长短期记忆和注意力机制的社交生成对抗网络用于行人轨迹预测[J]. 《计算机应用》唯一官方网站, 2023, 43(5): 1565-1570. |
[9] | 郭劲文, 马兴华, 骆功宁, 王玮, 曹阳, 王宽全. 基于Transformer的结构强化IVOCT导丝伪影去除方法[J]. 《计算机应用》唯一官方网站, 2023, 43(5): 1596-1605. |
[10] | 王昊, 王子成, 张超, 马韵升. 基于生成对抗网络的数据不确定性量化方法[J]. 《计算机应用》唯一官方网站, 2023, 43(4): 1094-1101. |
[11] | 樊小宇, 蔺素珍, 王彦博, 刘峰, 李大威. 基于残差图卷积神经网络的高倍欠采样核磁共振图像重建算法[J]. 《计算机应用》唯一官方网站, 2023, 43(4): 1261-1268. |
[12] | 尹春勇, 周立文. 基于再编码的无监督时间序列异常检测模型[J]. 《计算机应用》唯一官方网站, 2023, 43(3): 804-811. |
[13] | 陶玲玲, 刘波, 李文博, 何希平. 有闭解的可控人脸编辑算法[J]. 《计算机应用》唯一官方网站, 2023, 43(2): 601-607. |
[14] | 朱利安, 张鸿. 基于双分支条件生成对抗网络的非均匀图像去雾[J]. 《计算机应用》唯一官方网站, 2023, 43(2): 567-574. |
[15] | 陈刚, 廖永为, 杨振国, 刘文印. 基于多特征融合的多尺度生成对抗网络图像修复算法[J]. 《计算机应用》唯一官方网站, 2023, 43(2): 536-544. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||