EPCE-HDR (ECAI 2023 Oral)

This is a pytorch project for the paper High Dynamic Range Image Reconstruction via Deep Explicit Polynomial Curve Estimation

by Jiaqi Tang, Xiaogang Xu, Sixing Hu and Ying-Cong Chen* presented at ECAI 2023.

*: Corresponding author

Introduction

Due to different physical imaging parameters, the tone-mapping functions between images and real radiance are highly diverse, which makes HDR reconstruction extremely challenging. Existing solutions can not explicitly clarify a corresponding relationship between the tone-mapping function and the generated HDR image, but this relationship is vital when guiding the reconstruction of HDR images. To address this problem, we propose a method to explicitly estimate the tone mapping function and its corresponding HDR image in one network.

Paper link | Poster | Oral (PPT)

Updates

• ✅ Fix some bugs in codes.
• ✅ Fix the problem in datasets. Check this Google Cloud link for DOWNLOAD.
• ✅ Release Pretrained Model. Check this Google Cloud link for DOWNLOAD.

Getting Started

1. Multi TM Curves Dataset
2. Environment
3. Configuration
4. Usage
5. Evaluation

Multi TM Curves Dataset

• You can download the original dataset from Codalab.

https://competitions.codalab.org/competitions/28161#participate-get-data

• Or download it directly from Drive (Recommend): Google Drive. (including code for building dataset)

Environment

pip install -r requirements.txt

Config Files

codes
│
└───options
│   │
│   └───test
│   │   │ test.yml (for testing in Test-Real)
│   │   │ test_sys.yml  (for testing in Test-Gamma)
│   │    
│   └───train
│       │ train.yml  (for training)
│   

Usage

How to test

• Modify dataroot_LQ and pretrain_model_G for low dynamic range image and model, then run

cd codes
python test.py -opt options/test/test.yml
python test.py -opt options/test/test_sys.yml

The test results will be saved to ./results.

How to train

• Modify dataroot_LQ and dataroot_GT in ./codes/options/train/train, then run

cd codes
python train.py -opt options/train/train.yml

The models and training states will be saved to ./experiments.

Visualization

Modify paths of the data in ./scripts/tonemapped_visualization.py and run

cd scripts
python tonemapped_visualization.py

to visualize the images.

Evaluation

• In addition to the recognized evaluation metrics (PSNR, SSIM etc.), we also introduce AvgPSNR as an evaluation metric. AvgPSNR is a metric that can jointly assess the PSNR values of HDR images before and after tone mapping. It is defined as,

AvgPSNR = 0.7 * PSNR + 0.3 * mu-PSNR

• Testing in the wild

Citation Information

If you find the project useful, please cite:

@inproceedings{Tang2023HighDR,
  title={High Dynamic Range Image Reconstruction via Deep Explicit Polynomial Curve Estimation},
  author={Jiaqi Tang and Xiaogang Xu and Sixing Hu and Ying-Cong Chen},
  booktitle={European Conference on Artificial Intelligence},
  year={2023}
}

Contributions

If you have any questions, feel free to e-mail the author Jiaqi Tang (jtang092@connect.hkust-gz.edu.cn). Besides, these codes are built on HDRUNET.