A Meaningful Reformulation of Relative Spectral Discrimination Power to Analyze Hyperspectral Data

Abstract

Spectral matching algorithms (SMAs) discriminate and distinguish spectral signatures of earth surface features by comparing with their ground-truth spectra. Though different SMAs developed based on different theoretical strategies, choosing an effective SMA is still a challenging task. To study the performance of SMAs in distinguishing spectral signatures, few performance measure are developed and relative spectral discrimination power (RSDPW) is one such a measure. RSDPW discriminates how one spectral signature is distinct from another relative to a reference spectral signature. Classical way of measuring RSDPW do not takes into account of spectral matching between the two spectral signatures to be discriminated. Therefore, in this paper, a reformulation for RSDPW is presented to get a good idea about the spectral diversity of spectral signatures to measure RSDPW in a more meaningful manner and also to make it perspicacious. The experimental results show that the proposed reformulated RSDPW not only a meaningful way to measure it but also robust/standard enough to compare various SMAs by measuring it. Additionally, the range of RSDPW values for different levels of discrimination is demonstrated for the present study. © 2023 IEEE.