In this paper, we present a simple yet powerful deep neural network for natural image dehazing. The proposed method is designed based on U-Net architecture and we made some design changes to make it better. We first use Group Normalization to replace Batch Normalization to solve the problem of insufficient batch size due to hardware limitations. Second, we introduce FReLU activation into the U-Net block, which can achieve capturing complicated visual layouts with regular convolutions. Experimental results on public benchmarks demonstrate the effectiveness of the modified components. On the SOTS Indoor and Outdoor datasets, it obtains PSNR of 32.23 and 31.64 respectively, which are comparable performances with state-of-the-art methods. The code is publicly available online soon.
Wenyi GE
Chengdu University of Information Technology
Yi LIN
Sichuan University
Zhitao WANG
Beijing Satellite Navigation Center (BSNC)
Guigui WANG
Sichuan University
Shihan TAN
Sichuan University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Wenyi GE, Yi LIN, Zhitao WANG, Guigui WANG, Shihan TAN, "An Improved U-Net Architecture for Image Dehazing" in IEICE TRANSACTIONS on Information,
vol. E104-D, no. 12, pp. 2218-2225, December 2021, doi: 10.1587/transinf.2021EDP7043.
Abstract: In this paper, we present a simple yet powerful deep neural network for natural image dehazing. The proposed method is designed based on U-Net architecture and we made some design changes to make it better. We first use Group Normalization to replace Batch Normalization to solve the problem of insufficient batch size due to hardware limitations. Second, we introduce FReLU activation into the U-Net block, which can achieve capturing complicated visual layouts with regular convolutions. Experimental results on public benchmarks demonstrate the effectiveness of the modified components. On the SOTS Indoor and Outdoor datasets, it obtains PSNR of 32.23 and 31.64 respectively, which are comparable performances with state-of-the-art methods. The code is publicly available online soon.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2021EDP7043/_p
Copy
@ARTICLE{e104-d_12_2218,
author={Wenyi GE, Yi LIN, Zhitao WANG, Guigui WANG, Shihan TAN, },
journal={IEICE TRANSACTIONS on Information},
title={An Improved U-Net Architecture for Image Dehazing},
year={2021},
volume={E104-D},
number={12},
pages={2218-2225},
abstract={In this paper, we present a simple yet powerful deep neural network for natural image dehazing. The proposed method is designed based on U-Net architecture and we made some design changes to make it better. We first use Group Normalization to replace Batch Normalization to solve the problem of insufficient batch size due to hardware limitations. Second, we introduce FReLU activation into the U-Net block, which can achieve capturing complicated visual layouts with regular convolutions. Experimental results on public benchmarks demonstrate the effectiveness of the modified components. On the SOTS Indoor and Outdoor datasets, it obtains PSNR of 32.23 and 31.64 respectively, which are comparable performances with state-of-the-art methods. The code is publicly available online soon.},
keywords={},
doi={10.1587/transinf.2021EDP7043},
ISSN={1745-1361},
month={December},}
Copy
TY - JOUR
TI - An Improved U-Net Architecture for Image Dehazing
T2 - IEICE TRANSACTIONS on Information
SP - 2218
EP - 2225
AU - Wenyi GE
AU - Yi LIN
AU - Zhitao WANG
AU - Guigui WANG
AU - Shihan TAN
PY - 2021
DO - 10.1587/transinf.2021EDP7043
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E104-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2021
AB - In this paper, we present a simple yet powerful deep neural network for natural image dehazing. The proposed method is designed based on U-Net architecture and we made some design changes to make it better. We first use Group Normalization to replace Batch Normalization to solve the problem of insufficient batch size due to hardware limitations. Second, we introduce FReLU activation into the U-Net block, which can achieve capturing complicated visual layouts with regular convolutions. Experimental results on public benchmarks demonstrate the effectiveness of the modified components. On the SOTS Indoor and Outdoor datasets, it obtains PSNR of 32.23 and 31.64 respectively, which are comparable performances with state-of-the-art methods. The code is publicly available online soon.
ER -