An Improved Underwater Color Image Enhancement Algorithm Based on CycleGAN
-
摘要: 针对基于深度学习方法的水下图像增强只考虑水下图像的RGB 颜色特征空间造成的增强效果不理想现象,本文在循环生成对抗网络的基础上改进了一种水下彩色图像增强算法。首先运用循环生成对抗网络在图像的RGB和HSV 颜色特征空间进行训练,将图像经过卷积网络下采样提取到的特征送入残差网络和扩展压缩模块,其中扩展压缩模块可以调整图像RGB 和HSV 通道的权重。预训练好的生成对抗网络作用在成对的水下降质图像与增强后的图像进行监督训练,采用特征融合网络将对抗生成网络输出的RGB 和HSV 六通道图像融合成RGB三通道图像。实验结果表明,该方法能够有效结合图像的RGB和HSV 空间的特征信息,提升水下图像的对比度和亮度,校正水下图像的颜色偏差。Abstract: The underwater image enhancement based on the deep learning method considers only the RGB feature space, therefore the image enhancement effect is unsatisfactory. To cope with this problem, this paper proposed an improved underwater color image enhancement algorithm based on the cyclic generative adversarial network (CycleGAN). Both RGB and HSV color spaces of an image are used to train the CycleGAN. The features down-sampled from the CycleGAN are input into the residual network and the expansion compression module to extract useful features. The weights of RGB and HSV spaces are adaptively adjusted in the expansion and compression module. The pre-trained CycleGAN acts on the paired water degraded image and the enhanced image for weakly supervised training. The feature fusion network is adopted to fuse the output of the CycleGAN into three channels of a new RGB image. The experimental results show that the algorithm can effectively combine the feature information on both RGB and HSV spaces, improves the contrast and brightness of the underwater image and corrects its color deviation.
-
表 1 增强图像的指标对比
Table 1. Performance parameters for enhanced images with different image enhancing algorithms
方法 MSE/103 PSNR/dB SSIM Retinex-based 1.2924 17.0168 0.6071 Dense GAN 1.7298 15.7058 0.521 Water Cycle GAN 1.2152 17.2843 0.4426 Fusion-based 1.128 17.6077 0.7721 Ours 1.1681 18.4355 0.6808 -
[1] WANG Y, ZHANG J, CAO Y, et al. A deep CNN method for underwater image enhancement[C]//2017 IEEE International Conference on Image Processing (ICIP). Beijing: IEEE, 2017: 1382-1386. [2] LI C Y, GUO C L, REN W Q, 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 [3] LI J, SKINNER K A, EUSTICE R M, et al. WaterGAN: unsupervised generative network to enable real-time color correction of monocular underwater images[J]. IEEE Robotics and Automation Letters, 2018, 3(1): 387-394. [4] GUO Y C, LI H Y, ZHUANG P X. Underwater image enhancement using a multiscale dense generative adversarial network[J]. IEEE Journal of Oceanic Engineering, 2020, 45(3): 862-870. doi: 10.1109/JOE.2019.2911447 [5] LI C Y, GUO J C, GUO C L. Emerging from water: underwater image color correction based on weakly supervised color transfer[J]. IEEE Signal Processing Letters, 2018, 25(3): 323-327. doi: 10.1109/LSP.2018.2792050 [6] MEJJATI Y A, RICHARDT C, TOMPKIN J, et al. Unsupervised attention-guided image-to-image translation[C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems. Montréal: Curran Associates Inc. , 2018: 3697-3707. [7] 李庆忠, 白文秀, 牛炯. 基于改进CycleGAN的水下图像颜色校正与增强[J]. 自动化学报, 2023, 49(4): 820-829.LI Q Z, BAI W X, NIU J. Underwater image color correction and enhancement based on improved cycle-consistent generative adversarial networks[J]. Acta Automatica Sinica, 2023, 49(4): 820-829. (in Chinese) [8] 杨亚绒, 李恒, 赵磊, 等. 改进的同态滤波与多尺度融合的水下图像增强[J]. 机械科学与技术, 2022, 41(8): 1231-1239. doi: 10.13433/j.cnki.1003-8728.20200455YANG Y R, LI H, ZHAO L, et al. Improved homomorphic filtering and multi-scale fusion method for underwater image enhancement[J]. Mechanical Science and Technology for Aerospace Engineering, 2022, 41(8): 1231-1239. (in Chinese) doi: 10.13433/j.cnki.1003-8728.20200455 [9] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems. Montreal: MIT Press, 2014: 2672-2680. [10] 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. Venice: IEEE, 2017: 2242-2251. [11] ARJOVSKY M, CHINTALA S, BOTTOU L. Wasserstein GAN[J]. Machine Learning, 2017, 3: 1701-1732. [12] EDWARDS D A. On the Kantorovich-Rubinstein theorem[J]. Expositiones Mathematicae, 2011, 29(4): 387-398. doi: 10.1016/j.exmath.2011.06.005 [13] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large scale image recognition[J]. Computer Science, 2014, 1(6): 1-14. [14] FU X Y, ZHUANG P X, HUANG Y, et al. A retinex-based enhancing approach for single underwater image[C]//2014 IEEE International Conference on Image Processing (ICIP). Paris: IEEE, 2014: 4572-4576. [15] ANCUTI C, ANCUTI C O, HABER T, et al. Enhancing underwater images and videos by fusion[C]//2012 IEEE Conference on Computer Vision and Pattern Recognition. Providence: IEEE, 2012: 81-88.