Hyperspectral Data Processing: Algorithm Design and Analysis

26.1 Introduction

Spectral signature coding (SSC) is a scheme, a rule, or a mapping that transforms spectral values into a new set of symbols in a specific manner that a signature vector can be represented by the new symbols more effectively or efficiently. In hyperspectral data, each data sample is acquired by hundreds of spectral channels to form a column vector that can be used to diagnose subtle material substances based on their spectral characteristics. Therefore, taking advantage of such intraband spectral information (e.g., spectral information provided by spectral channels within a hyperspectral data sample vector) is one of the great benefits resulting from hyperspectral data. However, this also is traded off for a price that many unknown spectral signature vectors may be also extracted to further complicate spectral analysis. So, one of the major challenges in hyperspectral data exploitation is how to best utilize the spectral information provided by hyperspectral data to accomplish tasks such as detection, discrimination, classification, identification, while discarding undesired information caused by unwanted interference such as noise.

This chapter investigates a new approach to SSC, called progressive spectral signature coding (PSSC), where SSC is carried out in a progressive fashion rather than sequential coding by classical coding methods. It is a technique that can decompose a signature vector in multiple stages where each of these stages captures spectral changes ...