This is the code of paper 'DestripeCycleGAN: Stripe Simulation CycleGAN for Unsupervised Infrared Image Destriping'
DestripeCycleGAN: Stripe Simulation CycleGAN for Unsupervised Infrared Image Destriping. Shiqi Yang, Hanlin Qin, Shuai Yuan, Xiang Yan
The main contributions of this paper are as follows:
-
An efficient deep unsupervised DestripeCycleGAN is proposed for infrared image destriping. We incorporated a stripe generation model (SGM) into the framework, balancing the semantic information between the degraded and clean domains.
-
The Haar Wavelet Background Guidance Module (HBGM) is designed to mitigate the impact of vertical stripes and accurately assess the consistency of background details. As a plug-and-play image constraint module, it can offer a powerful unsupervised restriction for DestripeCycleGAN.
-
We design multi-level wavelet U-Net (MWUNet) that leverages Haar wavelet sampling to minimize feature loss. The network effectively integrates multi-scale features and strengthens long-range dependencies by using group fusion block (GFB) in skip connections.
If you find the code useful, please consider citing our paper using the following BibTeX entry.
@article{yang2024destripecyclegan,
title={DestripeCycleGAN: Stripe Simulation CycleGAN for Unsupervised Infrared Image Destriping},
author={Yang, Shiqi and Qin, Hanlin and Yuan, Shuai and Yan, Xiang and Rahmani, Hossein},
journal={arXiv preprint arXiv:2402.09101},
year={2024}
}
- Our project has the following structure:
python train.py
python test.py
Welcome to raise issues or email to [email protected] or [email protected] for any question regarding our DestripeCycleGAN.