xiexinheng/NODE-ImgNet

NODE-ImgNet: a PDE-informed effective and robust model for image denoising

This repository serves as the official test implementation of my paper, which has been preprinted on arXiv.

• Xie, X., Wu, Y., Ni, H. and He, C., 2023. NODE-ImgNet: a PDE-informed effective and robust model for image denoising. arXiv preprint arXiv:2305.11049.

The provided code in this repository corresponds to the concepts and methodologies described in the paper.

Network architecture

Running Codes

Requirements for a successful implementation of the codes can be found in requirements.txt.

Training datasets

The training dataset of Gaussian noisy is downloaded at https://kedema.org/project/exploration/index.html

After downloading, place the folder in the ./data/GaussianTrainingData/ directory.

Commands

Below are some example commands for training or testing NODE-ImgNet:

Training NODE-ImgNet for gray Gaussian noisy images with σ = 25

python gaussian_gray_denoising.py --noise_level 25

Training NODE-ImgNet for color Gaussian noisy images with σ = 25

python gaussian_color_denoising.py --noise_level 25

Training NODE-ImgNet-B (Blind) for gray Gaussian noisy images

python gaussian_gray_denoising.py --is_blind

Training NODE-ImgNet-B (Blind) for color Gaussian noisy images

python gaussian_color_denoising.py --is_blind

If you want to cite this paper, please refer to the following format

1. Xie, X., Wu, Y., Ni, H. and He, C., 2023. NODE-ImgNet: a PDE-informed effective and robust model for image denoising. arXiv preprint arXiv:2305.11049.

2. @article{xie2023node,

   title={NODE-ImgNet: a PDE-informed effective and robust model for image denoising},

   author={Xie, Xinheng and Wu, Yue and Ni, Hao and He, Cuiyu},

   journal={arXiv preprint arXiv:2305.11049},

   year={2023} }