Hyperspectral Data Processing: Algorithm Design and Analysis

24.1 Introduction

Binary coding is one of simplest coding methods to represent data with binary values, {0,1}. One such method is the bit plane coding commonly used in image compression where an image is represented in accordance with its bit significance in terms of gray-scale values (Gonzalez and Woods, 2002). The idea of applying binary coding to spectral data was first proposed by Mazer et al. (1988) who developed a binary coding-based image software system called SPAM for remotely sensed imagery. Since remotely sensed data samples are collected by a number of spectral channels simultaneously, a data sample is actually a column vector with its components made up of data samples acquired by separate spectral channels. More specifically, assume that L is the number of spectral channels used for data acquisition. Each data sample vector, referred to as signature vector, is represented by an L-dimensional vector with the lth signature component specified by the data sample in the lth spectral band. SPAM binary coding encodes an L-dimensional signature vector as a (2L-2)-dimensional binary code word composed of the first L binary values encoding the sign of the difference between a signature component sample and its spectral signature mean, and a new set of additional L-2 binary values encoding the sign of the difference between a component pixel in a signature vector and its adjacent signature components within the signature vector to be encoded. SPAM binary coding was further ...