SNRGAN: The Semi Noise Reduction GAN for Image Denoising

Author: [Mehrshad Momen-Tayefeh, Mehrdad Momen-Tayefeh, Amir Ali Ghafourian Ghahramani]
Published in: [International Symposium on Artificial Intelligence and Signal Processing (AISP) IEEE]
DOI: [10.1109/AISP61396.2024.10475264]

Description

This repository contains the implementation and resources for the paper “Semi Noise Reduction GAN (SNRGAN) for Efficient Noise Reduction”. Conventional noise reduction methods often struggle with high noise levels, leading to artifacts and distortions. In this work, we propose a low-complexity Generative Adversarial Network (GAN) approach for noise reduction, termed SNRGAN. Our model effectively learns noise patterns and generates denoised images across various noise levels.

The model was trained on three diverse datasets, achieving superior PSNR and NMSE scores compared to traditional methods. SNRGAN demonstrated strong performance in both subjective evaluations and objective metrics, making it a promising solution for real-world noise reduction applications.

Contents

• Code: Source code for implementing the methods described in the paper.
• Paper: https://ieeexplore.ieee.org/abstract/document/10475264.

Requirements

• Python >= V 3.10.12

Citation

If you find this work useful, please cite the paper:

@INPROCEEDINGS{10475264,
  author={Momen-Tayefeh, Mehrshad and Momen-Tayefeh, Mehrdad and Hasheminasab, Fatemeh Zahra and Ghahramani, S. AmirAli Gh.},
  booktitle={2024 20th CSI International Symposium on Artificial Intelligence and Signal Processing (AISP)}, 
  title={SNRGAN: The Semi Noise Reduction GAN for Image Denoising}, 
  year={2024},
  volume={},
  number={},
  pages={1-5},
  keywords={Training;Image quality;Quantization (signal);Noise reduction;Neural networks;Generative adversarial networks;Distortion;Generative Adversarial Networks;Noise Reduction;Convolutional Neural Networks},
  doi={10.1109/AISP61396.2024.10475264}}