Hyperspectral Data Processing: Algorithm Design and Analysis

26

Progressive Coding for Spectral Signatures

Spectral signature coding is an effective means of characterizing spectral features and is performed by encoding signature vectors sequentially. This chapter develops a rather different encoding concept called progressive signature coding (PSC) that encodes a signature vector in a progressive manner. More specifically, it progressively encodes a spectral signature vector in multiple stages, each of which captures different but disjoint spectral information contained in the spectral signature vector. As a result of such a progressive coding, a profile of progressive changes in spectral variation for a spectral signature vector can be generated for spectral characterization. The proposed idea is very simple and evolved from the pulse code modulation (PCM), which is a commonly used quantization technique in communications and signal processing. It expands PCM to multiple-stage PCM (MPCM) in the sense that a signature vector can be decomposed and quantized by PCM progressively in multiple stages for spectral characterization. In doing so, MPCM generates a priority code for a spectral signature vector so that its spectral information captured in different stages can be prioritized in accordance with significance of changes in spectral variation. Such a coding, referred to as MPCM-based progressive spectral signature coding (MPCM-PSSC), can be very useful in many applications in hyperspectral data exploitation.