Mean-removed Nearest Neighbor Reordering Based Lossless Compression of 3D Hyperspectral Sounder Data

Hyperspectral sounder data is used for retrieval of atmospheric temperature, moisture and trace gases profiles, surface temperature and emissivity, cloud and aerosol optical properties. The physical retrieval of these geophysical parameters is a mathematically ill-posed problem whose solution is sensitive to the error or noise in the data. Therefore, lossless or near lossless compression of hyperspectral sounder data is desired to avoid potential retrieval degradation of the geophysical parameters. In addition to the spatial correlations of observed nature scenes, the hyperspectral sounder data features high correlations in disjoint spectral regions affected by the same type of absorbing gases. A preprocessing scheme to explore the spectral and spatial correlations will be beneficial for compression gains. In this paper we investigate Mean-removed Nearest Neighbor Reordering (MR-NNR) for preprocessing the sounder data. The result is then encoded using stateof-the-art compression algorithms such as CALIC, JPEG-LS and JPEG2000. It is shown that by use of the MR-NNR scheme, the compression gains of CALIC, JPEG-LS and JPEG2000 increase up to 15%, 5% and 7% respectively over the original data without any preprocessing.