Advancements in Deep Learning-Based Super-resolution for Remote Sensing: A Comprehensive Review and Future Directions

The spatial and spectral resolution limitations of Earth observation satellites pose significant challenges in critical fields such as change monitoring and making image super-resolution (ISR) techniques essential for improving the quality, interpretability, and accuracy of remote sensing data. This chapter provides a comprehensive review of advancements in super-resolution (SR) techniques for remote sensing image enhancement, with a particular emphasis on deep learning (DL) approaches. Both single-image super-resolution (SISR) and multi-image super-resolution (MISR) methods were reviewed. The development of DL-based SISR algorithms that have substantially enhanced satellite image quality was discussed. The different SR techniques were categorized into interpolation-based, reconstruction-based, learning-based, convolutional neural network (CNN)-based, and transformer-based algorithms. Challenges such as the limitations of current remote sensing datasets and the discrepancies between low-resolution (LR) and high-resolution (HR) mappings were also discussed. By providing a thorough review of DL-based SISR and MISR methods, this review offers insights into recent advancements and future research directions in this field.