- Diego Antonio Villalba González
- Arantxa Camil Junco Flores
- Jonathan Alexis Hernández Cuevas
This project presents the development of a Generative Adversarial Network (GAN) combined with planetary image processing to enhance resolution, detail, and contrast from low-resolution and high-noise images.
We achieved satisfactory results, applicable to various image processing fields, such as security cameras, medical imaging (ultrasounds), and astronomy.
Astrophotography is a technique that combines astrophysics and photography to capture images of celestial objects. It has evolved from photographic plates to high-resolution digital cameras.
Technological advances have enabled shorter exposure times, real-time image visualization, and improved post-processing techniques. However, challenges such as light pollution, digital noise, and tracking stability of telescopes still impact image quality.
SRGANs are a type of deep learning model that consists of two competing neural networks: a generator and a discriminator. The generator takes a low-resolution image and attempts to create a high-resolution version, while the discriminator learns to differentiate between real and generated images.
By competing against each other, both networks improve, resulting in higher-quality image reconstruction.
Planetary astrophotography is particularly challenging because:
- Planets appear small and move quickly in the night sky.
- Atmospheric turbulence and motion blur reduce image quality.
- Typical photography resolution is limited by video capture quality.
A common solution is video stacking, where multiple video frames are combined to reduce noise and improve detail. However, this method is still constrained by video resolution limits.
Our project enhances planetary images by applying SRGANs, allowing for higher resolution and improved details even when starting with low-quality sources.
git clone https://github.com/your_username/your_repo.git
cd your_repo- Create a Virtual Environment
python -m venv venv
source venv/bin/activate # On Windows, use venv\\Scripts\\activate- Install Dependencies
pip install -r requirements.txt-
Prepare Your Data • Place low-resolution planetary images in data/low_res/. • If available, place high-resolution images in data/high_res/ for training.
-
Train the Model
python train.py --config configs/train_config.yamlModify the configuration file as needed for different datasets.
- Generate Super-Resolution Images
Once trained, use the model to enhance new images:
python generate.py --input data/low_res/ --output results/high_res/We welcome contributions!
- Fork the repository.
- Create a new branch (
git checkout -b feature-branch). - Make your changes.
- Commit your changes (
git commit -m 'Added feature X'). - Push to the branch (
git push origin feature-branch). - Open a Pull Request.
For major changes, please open an issue first to discuss them
This project is licensed under the MIT License. See the LICENSE file for details.
The main reference for this project is the following paper:
- Ledig, C., Theis, L., Huszár, F., Caballero, J., Cunningham, A., Acosta, A., Aitken, A., Tejani, A., Totz, J., Wang, Z., & Shi, W. (2017).
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network.
arXiv:1609.04802.
For further details, you can access the original paper here.
Contact
For any questions or collaboration opportunities: • Diego Villalba • Email: diego.villalba@ciencias.unam.mx • LinkedIn: Diego Antonio Villalba González • GitHub: DiegoViillalba