Compact data structures for remote sensing data

Abstract

In this digital era, an enormous amount of data are being generated and processed daily. They accumulate to such an extent that it is necessary and imperative to use data compression to reduce the data size so that they can take up as little space as possible. Among the many data compression schemes, there is one known as compact data structures, and it will be the focus of this thesis. These structures store data efficiently while also providing real-time access to the data in the compressed domain, i.e., to query an individual element, it is not necessary to decompress the whole structure. Compact data structures also provide lossless compression, thus ensuring no information loss during the compression process. Remote sensing hyperspectral scenes are image data that are transmitted from sensors located in aircraft or in satellites orbiting the Earth to receivers at ground stations. Due to the size of the data, they need to be compressed in such a way that they can be transmitted more quickly and when they reach the ground stations, they can be stored in an efficient manner to save space. Therefore, data compression is necessary for faster transmission and reduced storage space. This thesis sets out to explore several distinct ways of using compact data structures to provide better performance with regard to compression ratios and access time for remote sensing hyperspectral data. First, we describe a predictive method and a differential method designed to work with a compact data structure and evaluate the improvements made. Then we present a study of different variable-length codes that can be used in tandem with compact data structures to achieve higher compression gains. Next, we analyze the tree structure of the raster matrix so that only nodes that contain relevant data are saved, thus making the structure more compact. Finally, we investigate a recently proposed compact data structure and examine how its performance stacks up against the others. Experiments have shown that these proposed methods produce results that remain competitive with the traditional techniques and methods that have been in use.