Overview and Introduction

Abstract

Hyperspectral imaging has become an emerging technique in remote sensing and has also successfully found its way into many other applications such as medical imaging, medical care, health care, and food industries for grading, safety, and inspection. Its benefits and advantages come from its use of as many as hundreds of spectral bands with very high spatial/spectral resolution. However, it also pays a heavy price for the excessive data volumes needing to be processed. For example, in satellite communication this is a very challenging issue because of long distance transmission and limited available bandwidths as well as data storage. Also, in many real-world applications, real-time processing is important because decision making must be achieved on a timely basis. Despite the fact that real-time processing has been widely studied in recent years, it is unfortunate that most algorithms claiming to be real-time are actually not for the following reasons. First, theoretically speaking, there are no such real-time processes in practice because computer processing time is always required and causes time delay. Second, a real-time process must be causal in the sense that no data sample vectors beyond the current being processed data sample vector should be allowed to be included in the data processing. Third, a real-time process should take advantage of its processed information and only process so-called innovations information which is not available at the time the data processing takes place. Finally, many real-time processing algorithms currently being used assume that the data are collected after data acquisition and then process the collected data in a post-real-time fashion. So, technically speaking, these algorithms are not true real-time processes because they cannot be implemented in real time while the data are being collected at the same time. Accordingly, these algorithms cannot be used for real-time data communication and transmission. In recent applications hyperspectral imaging has the capability of finding targets that are generally not known by prior knowledge or identified by visual inspection, such as moving objects or instantaneous objects, which can only appear for a short time and may not reappear after they vanish. In this case, detecting these targets on a timely basis must be immediate and target detection must be carried out in a real-time fashion even when data are being collected during data acquisition. Unfortunately, many currently developed real-time processing algorithms generally do not meet these criteria and cannot be used for this purpose. This book takes up this task and is devoted to design and development of real-time processing algorithms for hyperspectral data processing from a perspective of Progressive HyperSpectral Imaging (PHSI).