Investigation of Predictor-based Schemes for Lossless Compression of 3D Hyperspectral Sounder Data

Hyperspectral sounder data is used for retrieval of surface properties and atmospheric temperature, moisture, trace gases, clouds and aerosols. This large volume three-dimensional data is taken from many scan lines containing cross-track footprints, each with thousands of infrared channels. Unlike hyperspectral imager data compression, hyperspectral sounder data compression is better to be lossless or near lossless in order not to substantially degrade the geophysical retrieval. In this paper, we review different prediction-based schemes including CALIC and JPEG-LS for hyperspectral sounder data compression. To exploit the high spectral correlations, we also develop a method for optimal spectral prediction in the least square sense. A comparison with the JPEG2000 wavelet-based scheme is presented. The results show that the developed optimal prediction scheme outperforms all the other schemes in terms of compression ratios.