Journal of Applied Optics, Volume. 45, Issue 5, 937(2024)
General progressive unsupervised image enhancement method focusing on value aware
Figures & Tables(11)
Fig. 1. General progressive unsupervised image enhancement method focusing on value aware
Fig. 4. Comparison of actual enhancement effect between proposed method and other enhancers on LOL test set
Fig. 5. Performance evaluation experiment of SiamAPN tracker with proposed method
Fig. 6. Visualization images of object tracking process and error curves for center position
Table 1. Detailed structure parameters of VAPEN model
View tableView in ArticleTable 1. Detailed structure parameters of VAPEN model
阶段 层 输入通道 输出通道 卷积核 激活函数 上/下采样 预编码 Conv0-1 1 16 3×3 - - Conv0-2 16 16 3×3 Sigmoid - 编码 Conv1-1 16 16 3×3 - 下采样 Conv1-2 16 16 3×3 Sigmoid - Conv2-1 16 16 3×3 - 下采样 Conv2-2 16 16 3×3 Sigmoid - Conv3-1 16 16 3×3 - 下采样 Conv3-2 16 16 3×3 Sigmoid - 解码 Conv1-1 32 16 3×3 - 上采样 Conv1-2 16 16 3×3 Sigmoid - Conv2-1 32 16 3×3 - 上采样 Conv2-2 16 16 3×3 Sigmoid - Conv3-1 32 16 3×3 - 上采样 Conv3-2 16 1 3×3 Tanh -
Table 2. Quantitative evaluation results on authoritative benchmark
View tableView in ArticleTable 2. Quantitative evaluation results on authoritative benchmark
方法 PSNR SSIM SCI_E 9.580 0.388 RUAS_U 11.309 0.483 RRDNet 11.378 0.519 DarkLighter 13.764 0.613 SCI_D 13.806 0.597 SCI_M 14.784 0.618 DCE 14.861 0.666 DCE++ 15.249 0.675 RUAS_D 16.146 0.578 RUAS_L 16.405 0.701 RetinexNet 16.774 0.539 LIME 16.920 0.620 本文方法 17.610 0.618
Table 3. Model parameters and runtime efficiency
ms View tableView in ArticleTable 3. Model parameters and runtime efficiency
ms 方法 SCI RetinexNet RRDNet DarkLighter DCE DCE++ RUAS LIME 本文方法 参数量 258 1358739 128167 74768 79416 10561 3438 - 35057 处理时间 0.55 66.5 16.6 12.52 12.38 2.08 17.16 94.4 4.45
Table 4. Ablation experiment
View tableView in ArticleTable 4. Ablation experiment
亮度约束项 平滑项 保真项 PSNR SSIM √ 13.995 0.471 √ 6.523 0.019 √ 7.553 0.159 √ √ 16.695 0.593 √ √ 13.994 0.472 √ √ 7.530 0.156 √ √ √ 17.610 0.618
Tools
Get Citation
Copy Citation Text
Aiguo ZHOU, Jilin ZHAO, Shan AN, Changhong FU. General progressive unsupervised image enhancement method focusing on value aware[J]. Journal of Applied Optics, 2024, 45(5): 937
Paper Information
Category:
Received: Mar. 26, 2024
Accepted: --
Published Online: Dec. 20, 2024
The Author Email: Jilin ZHAO (符长虹)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20