2020 Volume 14 Issue 2 Pages 138-146
Hyperspectral (HS) imagery provides 3D data that contains both spatial (two-dimensional) and spectral (one-dimensional) information acquired by spectroscopic imaging in a wide range of wavelengths from ultraviolet to near-infrared. HS images can visualize physical properties and phenomena that cannot be captured by the human eye or RGB cameras. However, capturing complete spatial―spectral information is often difficult because of the limitations of the optical design and/or measurement conditions. In addition, it is also difficult to avoid degradation due to various types of noise arising in the measurement process. In this paper, we review optimization-based techniques for effectively estimating the desired HS image from such incomplete and degraded measurement data.
