Improved self-calibration image enhancement algorithm based on attention mechanism and its application in maritime low-light images

Table 1. Experimental environment configuration
View table
View in Article
Table 1. Experimental environment configuration
配置名称配置参数
CPU 12× Xeon Platinum 8260
GPU NVIDIA GeForce RTX 3090
显存 24 G
深度学习框架 PyTorch1.8.0
CUDA NVIDIA CUDA 11.3

Table 2. Description of evaluation parameters

View table

View in Article

Table 2. Description of evaluation parameters

评价指标	说明
标准差	数值越大，图像质量越好
自然图像质量评价	数值越小，图像失真越小
平均梯度	数值越大，图像越清晰
信息熵	数值越大，图像信息越丰富

Table 3. Evaluation results of standard deviation

View table

View in Article

Table 3. Evaluation results of standard deviation

算法	标准差
算法	图像1	图像2	图像3	图像4	图像5	图像6
Retinex	21.23	21.57	28.75	38.61	24.33	22.06
RUAS	23.71	23.70	37.37	49.60	28.67	23.24
Retinex-Net	24.05	24.61	31.15	53.49	26.34	28.38
Zero-DCE	21.54	22.12	29.97	42.84	27.74	24.17
SCI	29.35	26.50	38.84	47.08	29.81	27.35
本文算法	37.75	32.45	47.70	55.99	30.34	31.57

Table 4. Evaluation results of natural image quality evaluator

View table

View in Article

Table 4. Evaluation results of natural image quality evaluator

算法	自然图像质量评价
算法	图像1	图像2	图像3	图像4	图像5	图像6
Retinex	10.73	10.61	10.73	11.61	10.36	13.08
RUAS	9.12	9.53	9.51	9.27	9.15	11.39
Retinex-Net	9.49	8.87	8.25	9.45	9.68	9.61
Zero-DCE	8.83	9.66	9.68	10.61	9.53	9.83
SCI	8.69	9.13	7.97	9.61	9.13	9.66
本文算法	8.11	8.55	7.56	8.81	8.37	8.62

Table 5. Evaluation results of average gradient

View table

View in Article

Table 5. Evaluation results of average gradient

算法	平均梯度
算法	图像1	图像2	图像3	图像4	图像5	图像6
Retinex	2.65	3.45	5.31	3.50	4.31	3.18
RUAS	6.53	5.96	7.31	6.46	5.59	4.23
Retinex-Net	5.92	6.48	7.95	7.14	4.13	6.35
Zero-DCE	6.76	5.85	6.86	5.43	4.68	5.72
SCI	7.02	6.58	11.57	9.93	6.82	6.44
本文算法	8.47	8.64	12.84	12.70	8.96	11.36

Table 6. Evaluation results of information entropy

View table

View in Article

Table 6. Evaluation results of information entropy

算法	信息熵
算法	图像1	图像2	图像3	图像4	图像5	图像6
Retinex	4.28	4.12	4.05	4.24	4.26	3.50
RUAS	4.52	4.45	4.62	5.21	5.09	4.17
Retinex-Net	4.71	4.39	4.77	5.18	4.53	4.86
Zero-DCE	4.35	4.20	4.52	4.51	4.46	4.25
SCI	4.75	4.48	4.91	5.65	5.27	4.98
本文算法	5.08	4.91	5.00	5.91	5.51	5.84

Table 7. Evaluation results of the testing datasets in terms of standard deviation
View table
View in Article
Table 7. Evaluation results of the testing datasets in terms of standard deviation
算法标准差
Exclusivey Dark LOL 自建海上数据集
Retinex 28.64 24.77 23.59
RUAS 35.98 30.63 24.37
Retinex-Net 33.18 28.35 27.86
Zero-DCE 31.24 29.71 26.64
SCI 39.36 35.19 29.24
本文算法 47.33 41.27 35.82

Table 8. Evaluation results of the testing datasets in terms of natural image quality evaluator
View table
View in Article
Table 8. Evaluation results of the testing datasets in terms of natural image quality evaluator
算法自然图像质量评价
Exclusivey Dark LOL 自建海上数据集
Retinex 11.09 14.29 10.68
RUAS 9.26 12.03 9.12
Retinex-Net 9.35 12.67 9.23
Zero-DCE 9.92 13.89 9.08
SCI 8.57 11.61 8.66
本文算法 8.04 9.86 8.12

Table 9. Evaluation results of the testing datasets in terms of average gradient
View table
View in Article
Table 9. Evaluation results of the testing datasets in terms of average gradient
算法平均梯度
Exclusivey Dark LOL 自建海上数据集
Retinex 5.68 4.96 4.03
RUAS 8.16 6.73 6.17
Retinex-Net 7.55 7.25 6.34
Zero-DCE 7.84 6.95 6.56
SCI 10.39 7.75 7.85
本文算法 14.65 9.08 8.86

Table 10. Evaluation results of the testing datasets in terms of information entropy
View table
View in Article
Table 10. Evaluation results of the testing datasets in terms of information entropy
算法信息熵
Exclusivey Dark LOL 自建海上数据集
Retinex 4.20 4.06 4.22
RUAS 4.82 4.32 4.49
Retinex-Net 4.88 4.91 4.60
Zero-DCE 4.64 4.73 4.31
SCI 5.15 4.93 4.68
本文算法 5.53 5.08 5.10

Table 11. Algorithm efficiency analysis
View table
View in Article
Table 11. Algorithm efficiency analysis
算法运行时间/s
Retinex 15.96
RUAS 16.63
Retinex-Net 17.92
Zero-DCE 16.47
SCI 15.17
本文算法 15.95

Table 12. Results of ablation experiment
View table
View in Article
Table 12. Results of ablation experiment
模型 LA LAM RN BCN AG IE
原SCI 8.66 4.92
SCI-1 √ √ √ 9.32 5.16
SCI-2 √ √ √ 9.18 5.07
SCI-3 √ √ √ 8.71 4.98
SCI-4 √ √ √ 10.72 5.19
本文算法 √ √ √ √ 10.86 5.24

Tools

Get Citation

Copy Citation Text

Li SU, Shihao CUI. Improved self-calibration image enhancement algorithm based on attention mechanism and its application in maritime low-light images[J]. Chinese Journal of Ship Research, 2025, 20(3): 339

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Weapon, Electronic and Information System

Received: Mar. 20, 2024

Accepted: --

Published Online: Jul. 15, 2025

The Author Email:

DOI:10.19693/j.issn.1673-3185.03833

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Experimental environment configuration

Table 1. Experimental environment configuration

Table 2. Description of evaluation parameters

Table 2. Description of evaluation parameters

Table 3. Evaluation results of standard deviation

Table 3. Evaluation results of standard deviation

Table 4. Evaluation results of natural image quality evaluator

Table 4. Evaluation results of natural image quality evaluator

Table 5. Evaluation results of average gradient

Table 5. Evaluation results of average gradient

Table 6. Evaluation results of information entropy

Table 6. Evaluation results of information entropy

Table 7. Evaluation results of the testing datasets in terms of standard deviation

Table 7. Evaluation results of the testing datasets in terms of standard deviation

Table 8. Evaluation results of the testing datasets in terms of natural image quality evaluator

Table 8. Evaluation results of the testing datasets in terms of natural image quality evaluator

Table 9. Evaluation results of the testing datasets in terms of average gradient

Table 9. Evaluation results of the testing datasets in terms of average gradient

Table 10. Evaluation results of the testing datasets in terms of information entropy

Table 10. Evaluation results of the testing datasets in terms of information entropy

Table 11. Algorithm efficiency analysis

Table 11. Algorithm efficiency analysis

Table 12. Results of ablation experiment

Table 12. Results of ablation experiment