A pytorch implementation of the "TFDL" and "DFTL" models in the "Joint Multi-Scale Tone Mapping and Denoising for HDR Image Enhancement" paper.
- Python 3.7
- Pytorch 1.11.0
- opencv
- torchvision
- cuda 10.2
- numpy
- matplotlib
- kornia
- rawpy
├── examples # Contains raw image examples from HDR+ dataset to test the model
├── results
│ └── TFDL # Test results will be saved here by default
├── snapshots # Pre-trained snapshots
│ └── TFDL.pth # Pre-trained TFDL model
├── csrnet.py
├── gaussian_pyramid.py
├── main_test.py # testing code
├── models.py # Our models are defined here
Note that snapshots/TFDL.pth is uploaded using git lfs, so please do git lfs fetch in order to download this file.
cd Joint-Multi-Scale-Tone-Mapping-and-Denoising-for-HDR-Image-Enhancement
python main_test.py
The script will process the images in "examples" folder and save the enhanced images to "results" folder.
The code is made available for academic research purpose only. This project is open sourced under MIT license.
@INPROCEEDINGS{9707563,
author = {Hu, Litao and Chen, Huaijin and Allebach, Jan P.},
title = {Joint Multi-Scale Tone Mapping and Denoising for HDR Image Enhancement},
booktitle = {2022 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)},
year = {2022},
volume = {},
number = {},
pages = {729-738},
doi={10.1109/WACVW54805.2022.00080}}
}
(Full paper: t.ly/TvIU or https://ieeexplore.ieee.org/document/9707563)
If you have any questions, please contact Litao Hu at hu430@purdue.edu.