# On training deep networks for satellite image super-resolution

## TL;DR

This paper investigates how the method of generating low-resolution training data affects the performance of deep learning models for satellite image super-resolution, highlighting the importance of data preparation for real-world applications.

## Contribution

It reveals that training data characteristics significantly impact super-resolution accuracy and suggests that improved data preparation routines are crucial for practical deployment.

## Key findings

- Training data generation method greatly influences SRR performance.
- Common bicubic downsampling may not be optimal for satellite images.
- Better data preparation can enhance real-world applicability of SRR.

## Abstract

The capabilities of super-resolution reconstruction (SRR)---techniques for enhancing image spatial resolution---have been recently improved significantly by the use of deep convolutional neural networks. Commonly, such networks are learned using huge training sets composed of original images alongside their low-resolution counterparts, obtained with bicubic downsampling. In this paper, we investigate how the SRR performance is influenced by the way such low-resolution training data are obtained, which has not been explored up to date. Our extensive experimental study indicates that the training data characteristics have a large impact on the reconstruction accuracy, and the widely-adopted approach is not the most effective for dealing with satellite images. Overall, we argue that developing better training data preparation routines may be pivotal in making SRR suitable for real-world applications.