Introduction to Super-Resolution for Remotely Sensed Hyperspectral Images

Super-resolution stands as one of the most prominent research areas in computer vision, aiming to augment the resolution of digital images. The majority of current state-of-the-art techniques rely on deep neural networks. While many of these are tailored for grayscale or natural color images, only a fraction are specifically designed for hyperspectral images, such as those captured by satellites. This chapter provides an overview of super-resolution methods designed for satellite hyperspectral imagery. Initially, the chapter outlines super-resolution for natural color images along with the most popular approaches, loss functions, hyperspectral datasets, and evaluation methods. Subsequently, the focus shifts to techniques specifically devised for hyperspectral imagery. These encompass single-image super-resolution, hyperspectral and multispectral image fusion, pansharpening, and multi-image super-resolution.