Hyperspectral bands prediction based on inter-band spectral correlation structure

Hyperspectral imaging has been widely studied in many applications; this type of sensors is imaging spectrometry sensors, which divide the waveband into hundreds of contiguous narrow bands for analysis. These images provide much richer and finer spectral information than traditional multispectral images do. Hyperspectral satellite images require enormous amounts of data space. NASA's satellite EO-1 (Earth Observing satellite) generates data at an unprecedented rate. It has been estimated to be over several gigabytes of data every day, most of it being hyperspectral image data. Given the large volume of hyperspectral image data that is generated each day, the use of robust data compression techniques will be beneficial to data transfer and archive as it a relatively new technique for remote sensing; compressing these data presents a challenge for the currently standardized general purpose compressors. Beside the three-dimensional nature of the data, which exhibit correlation along each dimension, individual samples have 16 bit or higher precision. Current state-of-the-art lossless compressors do not perform well on data sources with large dynamic range. Lossless compression is often required for data collection and archiving due to the cost of the acquisition and transmission process and due to the fact that original data may be needed for unforeseen analyses or further elaboration. In this paper, we analyze the spectral cross correlation between bands for AVIRIS and Hyperion hyperspectral data to investigate the possible criterion for measurement of highly correlated bands; we introduce a new technique to find best group of bands suitable for prediction based on spectral cross correlation matrix and weighted network of correlation.