Hyperspectral Band Selection Considering Spectral and Spatial Information

Feature extraction or selection (also known as band selection (BS) for hyperspectral images) can be used to create dimension reduction approaches for hyperspectral images. A hyperspectral image’s extracted features have minimal physical relevance; however, the chosen features (or bands) can maintain the original spectral information. As a result, band selection is increasingly frequently utilized for hyperspectral images. In this article, we offer three band selection approaches: Spectral angle mapper with spatial coherence (SAMSC), morphological profiles with spatial similarity (MPSS), and multi-feature similarity measure (MFSM). For band selection, our suggested approaches incorporate spectral and spatial information from hyperspectral images.