Hyperspectral Data Processing: Algorithm Design and Analysis

20

Progressive Spectral Dimensionality Process

Hyperspectral compression is considered the first step to preserve crucial and vital spectral information in the two-stage spectral/spatial compression proposed in Chapter 19, where dimensionality reduction by transform (DRT) and dimensionality reduction by band selection (DRBS) discussed in Chapter 6 play a vital role in dealing with the so-called curse of dimensionality in spectral compression. One key issue in implementing dimensionality reduction (DR) and DRBS for spectral compression is that the number of dimensions q to be retained after DRT and the number of bands to be retained after DRBS must be known a priori. Despite the fact that these two values, q and , can be estimated by virtual dimensionality (VD) developed in Chapter 5, VD is not a one-size-fit-all universal criterion for various applications. In order to mitigate the dependence on VD, this chapter develops a new DRT, to be called progressive spectral dimensionality process (PSDP), which introduces a new concept of dimensionality prioritization (DP) that revolutionizes how the commonly used DR is implemented. The transform used to perform DR (i.e., DRT) is a linear transformation that converts the original data dimensions into spectral transformed components, each of ...