Super-Resolution of Sentinel-2 Imagery

The Earth observation community is always looking for satellite images with higher spatial resolution. Earth observation satellites provide invaluable data for monitoring environmental changes, urban development, agricultural practices, and more. However, factors such as orbital distance, sensor limitations, and cost constraints often restrict the achievable spatial resolution. High-resolution imagery can provide more detailed, accurate insights, but these high-resolution sensors are costly, and frequently capturing imagery at such resolutions is not always feasible.  Super-resolution offers a way to overcome these limitations.

A simple up-scaling us bilinear or bicubic interpolation do not provide good results. Model based up-scaling can help. Deep learning models offer several advantages over interpolation techniques, for increasing the spatial resolution of remote sensing images, even if the differences might not always be immediately apparent.

1. Feature Preservation: Deep learning models are better at preserving intricate features and textures in images. They can learn complex patterns and structures, which helps in maintaining the quality of fine details.

2. Edge Sharpness: These models tend to produce sharper edges compared to bicubic interpolation, which can result in smoother and more visually appealing images.

3. Adaptive Learning: Deep learning models can adapt to different types of images and learn from a large dataset, making them more versatile and capable of handling various image characteristics.

While the differences might not always be significant in every scenario, especially if the images are not highly complex, these advantages can be crucial for specific applications in remote sensing where details and accuracy are paramount.

Super-Resolution Model

This implementation uses a single image super-resolution model trained in the frame of the EVOLAND Horizon Europe project. The network has been trained using the Sen2Venµs dataset, complimented with patches for B11 and B12. On GitHub you can find a Python implementation by Julien Michel, which is using this model. But this only supports standard Level 1C and Theia L2A images.

S2 Super-Resolution Plugin

This SNAP operator super-resolves the 10- and 20- meter bands, namely B2,B3,B4,B5,B6,B7,B8,B8A,B11,B12. In addition, the 60-meter bands, can be included. But they are only resampled using a bicubic interpolation. In contrast to the Python implementation, this plugin can also resample the geometry bands and all other bands and masks.

Enhance Processing with GPU

You can install the NVIDIA CUDA software package. This allows the use of the graphics card for the processing. This drastically improves performance by an order of magnitude.

More general information about this release can be found in the other blog:

Release of EOMasters Toolbox Pro v1.2

The EOMasters Toolbox Pro is available in the EOMasters Shop. You can activate a trial version when installing the plugin in SNAP. Simply request the trial key and decide after the trial period if you want to purchase it.

Tschüss & Goodbye

Marco