Hyperspectral Data Processing: Algorithm Design and Analysis

19

Exploitation-Based Hyperspectral Data Compression

A key to success in data compression is determined by how much information needed to be stored that will be later retrieved for future data processing. Whether or not a compression technique is effective is measured by the degree of loss in future information retrieval rather than information recovery. Accordingly, data loss does not necessarily mean information loss, which implies that data compression does not necessarily perform information compression. This is particularly evidential in hyperspectral data compression in which many research efforts have been directed to data compression rather than information compression. This chapter introduces a new concept of hyperspectral information compression and explores its utility in hyperspectral data exploitation. In particular, a three-stage process is developed to address issues arising in hyperspectral information compression where the first stage is spectral compression performed by spectral dimensionality reduction via interband de-correlation followed by two stages that implement either exploitation-based information compression in second stage and 3D data compression in the third stage or in a reverse order. Such three-stage hyperspectral information compression is measured by two key factors: information loss and exploitation-based compression criteria. In order to demonstrate the difference between data compression and information compression in hyperspectral image processing, ...