Optimizing 4D Lookup Table for Low-light Video Enhancement via Wavelet Priori.[WaveLUT]
WaveLUT optimizes the 4D lookup table (LUT) technique by wavelet prior, which effectively improves the accuracy of color mapping while maintaining the efficiency. Meanwhile, we explore a dynamic fusion strategy to effectively fuse different a priori knowledge. And we utilize multimodal semantics combined with Fourier frequency domain for perception-oriented appearance driving. Experimental results on benchmark datasets show that our approach reaches the state-of-the-art.
conda activate WaveLUT
pip install -r requirements
python models/WaveLUT/transformation/setup.py install
- Pre-Trained Models
- Test
- Train
SDSD and SMID datasets Wang. DID datasets Fu
download this model and put it into WaveLUT/ckpt/.
python evaluation.py -opt [YOUR_yml]
We plan to release the training code in an upcoming update.
@article{he2024optimizing,
title={Optimizing 4D Lookup Table for Low-light Video Enhancement via Wavelet Priori},
author={He, Jinhong and Xue, Minglong and Wang, Wenhai and Zhou, Mingliang},
journal={arXiv preprint arXiv:2409.08585},
year={2024}
}