HG-MTFE

Jaemin Park, An Gia Vien, Thuy Thi Pham, Hanul Kim, and Chul Lee

Official pytorch implementation for "Image Enhancement Based on Histogram-Guided Multiple Transformation Function Estimation," accepted for publication in IEEE Transactions on Consumer Electronics.
Paper

  

Preparation

Training data: Download from GoogleDrive

The ZIP file contains three test datasets:

• LOL dataset: 485 image pairs
• FiveK dataset: 4,500 image pairs
• EUVP dataset: 11,435 image pairs

Testing samples: Download from GoogleDrive

The ZIP file contains three test datasets:

• LOL dataset: 15 image pairs
• FiveK dataset: 500 image pairs
• EUVP dataset: 515 image pairs

Pretrained weights: Download from GoogleDrive

The ZIP file contains weight files trained with each training dataset.

Training

1. Put low-quality images of training dataset in ./data/train_data/input
2. Put high-quality images of training dataset in ./data/train_data/gt
3. Put test images in ./data/test_data/LOL
4. Put ground-truths of test images in ./data/test_gt
5. Run below commend:

python lowlight_train.py

6. The trained model is saved at ./models
7. The result images are saved at ./data/analysis

Testing

1. Put test images in ./data/test_data/LOL
2. Put ground-truths of test images in ./data/test_gt
3. Run below commend:

python lowlight_test.py

4. The result images are saved at ./data/analysis

Citation (To be updated)

If you find this work useful for your research, please consider citing our paper:

@article{Park2024HGMTFE,
    author={{Park, Jaemin and Vien, An Gia and Pham, Thuy Thi and Kim, Hanul and Lee, Chul}},
    booktitle={IEEE Transactions on Consumer Electronics},
    title={Image Enhancement Based on Histogram-Guided Multiple Transformation Function Estimation}, 
    year={},
    volume={},
    number={},
    pages={},
    doi={}}
}

License

See MIT License