EARHD

《Non-homogeneous Image Dehazing with Edge Attention Based on Relative Haze Density》

Abstract. Image dehazing is a widely used technology for recovering clear images from hazy inputs. However, most dehazing methods are designed to target a specific haze concentration, without considering the varying degrees of image degradation. Removing non-homogeneous haze from real-world images is challenging. To address this issue, this study proposes a dual-cycle framework based on relative haze density, in which inputs are regarded as both hazy images to be recovered by a restoration network (RNet) and clear images to be deteriorated by a degradation network (DNet). Edge attention blocks and multi-order derivative loss are proposed for RNet to enhance the details and colors. Furthermore, two multi-class discriminators are designed to distinguish between relative levels of haze density. Extensive experiments on both real-world and synthetic datasets demonstrate that the proposed method is superior to state-of-the-art approaches for non-homogeneous image dehazing using either supervised or unsupervised learning. This code is available at https://github.com/lizhangray/EARHD.

Environment

All environment configurations are provided in the "sym.yaml" file. Please refer to this file for setting up your environment.

Create and activate a new environment using the sym.yaml file:

conda env create -f sym.yaml
conda activate <environment_name>

Replace <environment_name> with the name specified in the sym.yaml file.

Checkpoints & Datasets

Res2Net Pretrain Model

[BaiduNetdisk]

NHHAZE2020 Model

This is a PTH file trained on the NHHAZE2020 dataset. Below is the download link from Baidu Netdisk and instructions for use. [BaiduNetdisk]

Datasets

You can download the NTIRE dataset from the following link: [NTIRE]

Directory Structure

Here is a simple diagram illustrating the directory structure:

project_root/
├── input/
├── weights/
│   └── res2net.pth
├── output/
│   └── bestpsnr.pth
├── sym.yaml
└── option.py

Run

# test
python main.py

Example

NHHAZE20 Image with 2x Downsampling

NHHAZE20 Image with 4x Downsampling

NHHAZE21 Image with 2x Downsampling

NHHAZE21 Image with 4x Downsampling

Citation

Please cite this paper in your publications if it is helpful for your tasks.

Deng, R. et al. (2024). Non-homogeneous Image Dehazing with Edge Attention Based on Relative Haze Density. In: Advanced Intelligent Computing Technology and Applications. ICIC 2024. Lecture Notes in Computer Science, vol 14868, pp. 15-28, Springer. https://doi.org/10.1007/978-981-97-5600-1_2

@InProceedings{deng2024nonhomogeneous,
    author    = {Deng, Ruting and Li, Zhan and Deng, Yifan and Long, Hang and Chen, Zhanglu and Kang, Zhiqing and Qiu, Zhichao},
    title     = {Non-homogeneous Image Dehazing with Edge Attention Based on Relative Haze Density},
    booktitle = {Proceedings of the International Conference on Intelligent Computing (ICIC)},
    year      = {2024},
    volume    = {14868},
    pages     = {15--28},
    publisher = {Springer},
}