Abstract

Hyperspectral image analysis has considerably evolved during the past decades. The conventional model-based image processing and machine learning techniques are not efficient for hyperspectral image analysis, therefore, other advanced models such as spatial- spectral models were proposed to boost the hyperspectral analysis. Recent advances in ma- chine learning i.e., deep learning (DL) confirm that if an adequate amount of training data is supplied then DL-based algorithms outperform the conventional (shallow) ones. However, the available training data is often limited in hyperspectral imaging, therefore, the advan- tage of DL-based algorithms compared with shallow ones remains an open question. In this paper, we address this issue for two vibrant fields in hyperspectral image analysis i.e., Unmixing and Feature Extraction for Classification.

© 2021 The Author(s)