Hyperspectral Data Processing: Algorithm Design and Analysis

24

Binary Coding For Spectral Signatures

Binary coding is the simplest way to characterize spectral features. One commonly used method is a binary coding-based image software system, called spectral analysis manager (SPAM) for remotely sensed imagery developed by Mazer et al. (1988). It makes use of spectral mean and interband spectral difference as thresholds to generate a binary code word for a spectral signature vector. It is generally effective and also very simple to implement. The SPAM binary coding was further extended to a spectral feature binary coding (SFBC) developed by Qian et al. (1996) by incorporating an additional binary code word produced by thresholding the spectral mean deviation to further improve the SPAM performance. This chapter revisits these two approaches and further develops three new binary coding methods, median partition (MP), halfway partition (HP), and equal probability partition (EPP), all of which can be implemented in conjunction with the ideas of SPAM and SFBC to create new sets of binary code words for spectral signature coding. As a result, different combinations of various binary coding methods can be used for spectral signature coding and applications in spectral discrimination and identification. Finally, a new criterion called a posteriori discrimination probability (APDP) is also introduced as a performance measure to coompare two different binray coding methods for performance analysis.