Progressive Band Processing for Hyperspectral Imaging

Hyperspectral imaging has emerged as an image processing technique in many applications. The reason that hyperspectral data is called hyperspectral is mainly because the massive amount of information provided by the hundreds of spectral bands that can be used for data analysis. However, due to very high band-to-band correlation much information may be also redundant. Consequently, how to effectively and best utilize such rich spectral information becomes very challenging. One general approach is data dimensionality reduction which can be performed by data compression techniques, such as data transforms, and data reduction techniques, such as band selection. This dissertation presents a new area in hyperspectral imaging, to be called progressive hyperspectral imaging, which has not been explored in the past. Specifically, it derives a new theory, called Progressive Band Processing (PBP) of hyperspectral data that can significantly reduce computing time and can also be realized in real-time. It is particularly suited for application areas such as hyperspectral data communications and transmission where data can be communicated and transmitted progressively through spectral or satellite channels with limited data storage. Most importantly, PBP allows users to screen preliminary results before deciding to continue with processing the complete data set. These advantages benefit users of hyperspectral data by reducing processing time and increasing the timeliness of crucial decisions made based on the data such as identifying key intelligence information when a required response time is short