Optics and Precision Engineering, Volume. 31, Issue 18, 2687(2023)
Global and local feature fusion image dehazing
Figures & Tables(13)
Table 1. Parameter information of the generator
View tableView in ArticleTable 1. Parameter information of the generator
网络层 输出
分辨率
输出维度 操作 输入图像 256×256 3 —— Conv+BN+GELU 256×256 64 3×3 conv MaxPool 128×128 64 2×2 maxpool Conv+BN+GELU 128×128 128 3×3 conv MaxPool 64×64 128 2×2 maxpool Conv+BN+GELU 64×64 256 3×3 conv MaxPool 32×32 256 2×2 maxpool Conv+BN+GELU 32×32 512 3×3 conv GLFFM 32×32 512 详见2.3节 Pixel Shuffle 64×64 256 2倍上采样 Pixel Shuffle 128×128 128 2倍上采样 Pixel Shuffle 256×256 64 2倍上采样 Enhancer 256×256 128 详见2.4节 Conv+Tanh 256×256 3 3×3 conv
Table 2. Objective evaluation results on synthetic image datasets
View tableView in ArticleTable 2. Objective evaluation results on synthetic image datasets
去雾算法 PSNR SSIM DCP 17.692 1 0.877 8 CAP 18.475 8 0.820 4 AODNet 19.784 2 0.869 8 EPDN 21.503 6 0.879 2 Pix2pix 27.153 4 0.929 3 FFA-Net 31.126 5 0.958 3 LD-Net 24.765 3 0.917 2 Ours 33.190 2 0.977 0
Table 3. Objective evaluation results on real image datasets
View tableView in ArticleTable 3. Objective evaluation results on real image datasets
去雾算法 PSNR SSIM DCP 17.982 3 0.698 2 CAP 17.104 6 0.656 4 AODNet 17.119 3 0.613 2 EPDN 17.216 8 0.708 1 Pix2pix 18.109 2 0.710 2 FFA-Net 17.658 2 0.691 0 LD-Net 17.132 8 0.632 5 Ours 19.311 5 0.747 8
Table 4. Objective evaluation results of different positional encoding methods
View tableView in ArticleTable 4. Objective evaluation results of different positional encoding methods
编码方式 合成图像数据集 真实图像数据集 PSNR SSIM PSNR SSIM 全局位置编码生成器 33.190 2 0.977 0 19.315 5 0.747 8 局部位置编码生成器 32.677 6 0.971 5 19.073 9 0.727 9 固定位置编码 32.698 2 0.972 5 18.977 3 0.724 7 相对位置编码 31.814 8 0.966 8 18.705 8 0.711 7 可学习位置编码 32.679 2 0.970 7 18.714 6 0.720 3 无位置编码 31.498 5 0.958 6 18.668 5 0.708 5
Table 5. Objective evaluation results of the enhancer
View tableView in ArticleTable 5. Objective evaluation results of the enhancer
是否包含
增强模块
合成图像数据集 真实图像数据集 PSNR SSIM PSNR SSIM 是 33.190 2 0.977 0 19.311 5 0.747 8 否 32.486 9 0.970 1 17.534 2 0.662 7
Tools
Get Citation
Copy Citation Text
Xin JIANG, Haitao NIE, Ming ZHU. Global and local feature fusion image dehazing[J]. Optics and Precision Engineering, 2023, 31(18): 2687
Paper Information
Category: Information Sciences
Received: Feb. 13, 2022
Accepted: --
Published Online: Oct. 12, 2023
The Author Email: JIANG Xin (xinjiang@zju.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20