Terahertz Image Enhancement Based on Generative Adversarial Network
-
摘要: 太赫兹扫描成像中,由于激光器功率波动和仪器振动等原因,导致图像对比度较低,成像质量有待提高,且目前针对太赫兹图像的处理还停留在传统算法阶段。本文结合深度学习思想,提出了一种基于生成式对抗网络的图像增强方法。通过对训练集图像引入模糊和噪声,学习低质量图像和高质量图像之间的映射关系,并将其应用在真实太赫兹图像中。实验结果表明,与双边滤波、非局部均值滤波等传统算法相比,本文方法可在改善图像细节的基础上显著提高图像对比度,且视觉体验良好,这为太赫兹图像增强提供了新思路。Abstract: In terahertz scanning imaging, the image contrast is low due to laser power fluctuation and instrument vibration, and the imaging quality needs to be improved. At present, the processing of terahertz image is still in the traditional algorithm stage. In this paper, an image enhancement method based on Generative Adversarial Network is proposed, which includes the idea of deep learning. By introducing blur and noise into the training set image, the mapping relationship between low-quality images and high-quality images is learned and applied to real terahertz images. The experimental results show that, compared with traditional algorithms such as bilateral filtering and non-local mean filtering, this method can significantly improve the image contrast on the basis of improving image details, and has a good visual sense, which provides a new idea for terahertz image enhancement.
-
Key words:
- terahertz image /
- neural network /
- image enhancement /
- image contrast
-
表 1 不同方法PSNR、对比度计算结果
Table 1. PSNR and contrast calculation results by differentmethods
Bilateral filtering Nonlocal mean filtering Our algorithm PSNR 28.18 26.87 25.28 Contrast 178.69 134.48 375.98 -
[1] 郑显华, 王新柯, 孙文峰, 等. 太赫兹数字全息术的研发与应用[J]. 中国激光, 2014, 41(2): 0209003. https://www.cnki.com.cn/Article/CJFDTOTAL-JJZZ201402004.htmZHENG Xianhua, WANG Xinke, SUN Wenfeng, et al. Developments and applications of the terahertz digital holography[J]. Chinese J. Lasers, 2014, 41(2): 0209003. https://www.cnki.com.cn/Article/CJFDTOTAL-JJZZ201402004.htm [2] 张馨, 赵源萌, 张存林. 被动式太赫兹图像分割算法[J]. 强激光与粒子束, 2013, 25(6): 1597-1600. https://www.cnki.com.cn/Article/CJFDTOTAL-QJGY201306056.htmZHANG Xin, ZHAO Yuanmeng, ZHANG Cunlin. Passive terahertz image segmentation algorithm[J]. High Power Laser and Particle Beams, 2013, 25(6): 1597-1600. https://www.cnki.com.cn/Article/CJFDTOTAL-QJGY201306056.htm [3] 徐利民. 高分辨太赫兹图像处理[D]. 西安: 中国科学院研究生院(西安光学精密机械研究所), 2013.XU Limin. High Resolution Terahertz Image Processing[D]. Xi an: University of Chinese Academy of Sciences (Xi'an Institute of Optics and Precision Mechanics), 2013. [4] 李琦, 杨永发, 胡佳琦. 一种用于太赫兹共焦扫描图像复原的复合算法[J]. 红外与激光工程, 2015, 44(1): 321-326. doi: 10.3969/j.issn.1007-2276.2015.01.055LI Qi, YANG Yongfa, HU Jiaqi. A composite algorithm used for terahertz confocal scanning image restoration[J]. Infrared and Laser Engineering, 2015, 44(1): 321-326. doi: 10.3969/j.issn.1007-2276.2015.01.055 [5] 杨永发, 李琦. 双边滤波算法的太赫兹共焦扫描图像去噪应用[J]. 激光与光电子学进展, 2015, 52(12): 121101. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ201512015.htmYANG Yongfa, LI Qi. Application of bilateral filtering algorithm on terahertz confocal scanning image denoising[J]. Laser & Optoelectronics Progress, 2015, 52(12): 121101. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ201512015.htm [6] DONG C, CHEN C L, HE K, et al. Learning a Deep Convolutional Network for Image Super-Resolution[M]//Computer Vision – ECCV Springer International Publishing, 2014: 184-199. [7] Ledig C, Theis L, Huszar F, et al. Photo-realistic single image super-resolution using a generative adversarial network[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 2017: 4681-4690. [8] 卢贺洋, 苏胜君, 袁明辉, 等. 太赫兹图像的超分辨率重建[J]. 红外技术, 2019, 41(1): 59-63. http://hwjs.nvir.cn/article/id/hwjs201901009LU Heyang, SU Shengjun, YUN Minghui, et al. Super-resolution reconstruction of terahertz image[J]. Infrared Technology, 2019, 41(1): 59-63. http://hwjs.nvir.cn/article/id/hwjs201901009 [9] Johnson J, Alahi A, LI F. Perceptual losses for real-time style transfer and super-resolution[C]//In European Conference on Computer Vision (ECCV), Springer, 2016: 694-711. [10] Gross S, Wilber M. Training and investigating residual nets, online[EB/OL]. [2016-02-04]. http://torch.ch/blog/. [11] Shi W, Caballero J, Huszar F, et al. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network[C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 1874-1883. [12] Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks[C]// International Conference on Learning Representations (ICLR), 2016: 1-16. [13] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[C]//International Conference on Learning Representations (ICLR), 2015: 268-282. [14] Bruna J, Sprechmann P, Lecun Y. Super-resolution with deep convolutional sufficient statistics[C]//International Conference on Learning Representations (ICLR), 2016: 352-369. [15] Gatys L A, Ecker A S, Bethge M. Texture synthesis using convolutional neural networks[C]//Advances in Neural Information Processing Systems (NIPS), 2015: 262-270. [16] Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets[C]//Advances in Neural Information Processing Systems (NIPS), 2014: 2672-2680.