Optics and Precision Engineering, Volume. 33, Issue 2, 298(2025)
Underwater image enhancement by integrating domain transfer and attention mechanisms
Figures & Tables(11)
Fig. 2. Image generation network structure based on domain transfer
Fig. 5. Comparison of qualitative experimental results of different methods on UIEB dataset
Fig. 6. Comparison of qualitative experimental results of different methods on EUVP dataset
Table 1. Comparison of quantitative experimental results of different methods on UIEB dataset
View tableView in ArticleTable 1. Comparison of quantitative experimental results of different methods on UIEB dataset
方法 UICM UISM UIConM UIQM UCIQE UDCP 6.037 3 5.573 2 0.103 6 2.186 4 0.557 5 MIP 5.909 2 4.959 1 0.160 0 2. 2 031 0.569 8 UWCNN 4.281 6 5.604 7 0.302 5 2.857 3 0.482 1 UWGAN 6.232 2 6.453 1 0.254 7 2.992 0 0.583 8 CWR 6.463 2 6.763 2 0.245 9 3.058 6 0.581 7 U-Shape 6.304 7 6.607 9 0.259 3 3.056 2 0.573 1 本文方法 6.633 1 6.824 5 0.262 3 3.140 1 0.602 1
Table 2. Comparison of quantitative experimental results of different methods on EUVP dataset
View tableView in ArticleTable 2. Comparison of quantitative experimental results of different methods on EUVP dataset
方法 UICM UISM UIConM UIQM UCIQE UDCP 6.524 5 4.933 0 0.151 3 2.181 6 0.599 0 MIP 6.075 1 4.474 1 0.157 4 2.055 3 0.607 5 UWCNN 6.321 3 5.810 7 0.267 4 2.850 2 0.530 2 UWGAN 6.412 8 6.491 3 0.241 0 2.959 4 0.547 7 CWR 6.550 9 6.791 0 0.237 9 3.040 7 0.608 5 U-Shape 6.401 1 6.561 7 0.251 1 3.015 9 0.573 5 本文方法 6.802 1 6.821 4 0.243 5 3.076 8 0.612 4
Table 3. Analysis of model complexity
View tableView in ArticleTable 3. Analysis of model complexity
方法 Params/M FLOPs/G FPS UDCP - - 0.5 MIP - - 0.6 UWCNN 0.4 1.9 14.4 UWGAN 1.9 12.5 17.8 CWR 11.4 42.4 15.3 U-Shape 65.5 66.2 11.2 本文方法 32.3 33.9 15.8
Table 4. Experimental results of different loss weight combinations
View tableView in ArticleTable 4. Experimental results of different loss weight combinations
UIQM UICQE 0.5 0.2 0.1 3.085 4 0.594 7 0.5 0.3 0.1 3.140 1 0.602 1 0.5 0.4 0.1 3.071 6 0.584 5 0.5 0.3 0.2 3.117 5 0.599 4
Table 5. Results of ablation experiment
View tableView in ArticleTable 5. Results of ablation experiment
模型 a b c d e 域迁移生成网络 √ √ √ 混合注意力编码模块 √ √ 全局域关联一致性损失 √ √ UIQM 2.796 3 3.035 4 2.875 9 3.123 5 3.140 1 UICQE 0.481 2 0.581 5 0.507 5 0.593 2 0.602 1
Tools
Get Citation
Copy Citation Text
Tingting YAO, Zihao FENG, Hengxin ZHAO. Underwater image enhancement by integrating domain transfer and attention mechanisms[J]. Optics and Precision Engineering, 2025, 33(2): 298
Paper Information
Category:
Received: Aug. 27, 2024
Accepted: --
Published Online: Apr. 30, 2025
The Author Email: Tingting YAO (ytt1030@dlmu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20