Hyperspectral Data Processing: Algorithm Design and Analysis

23.8 Conclusions

Progressive band selection (PBS) is a new theory developed for band selection (BS), which is particularly useful in data communication and transmission. It can be considered as variable dimensionality BS (VDBS) as opposed to FDBS that implements BS dynamically by making band dimensionality adapt to various applications. It takes advantage of progressive band dimensionality process (PBDP), developed in Chapter 21, via band prioritization (BP) and band de-correlation (BD) to fine-tune BS in a progressive manner so that the band previously selected can be either expanded or removed to accommodate practical constraints. It offers several benefits over the traditional fixed band dimensionality BS. First, PBS does not need to have the precise knowledge of the number of bands needed to be selected because it makes the number of selected bands, p, vary and selects bands progressively in a forward and backward manner without repeatedly implementing BS. Second, PBS can be implemented by tuning variable bands according to its applications, such as selecting different ranges of wavelengths with different bands. Third, PBS keeps track of previously selected bands to update results in band expansion and reduction to accomplish various tasks, such as data compression, communication, storage, archiving and transmission, and so on. Finally, PBS can be implemented in causality and real time, a new emerging area, called progressive band processing in hyperspectral imaging, which ...