Adaptive quantized PCA with 3D prediction and positive packing for lossless compression of ultraspectral sounder data

Abstract

Given the unprecedented size of three-dimensional ultraspectral sounder data with high spectral resolution, lossless compression is preferable to avoid substantial degradation of the geophysical retrieval. A lossless compression method for ultraspectral sounder data is therefore developed. A quantized-principal-component-analysis-based scheme is presented by combining 3D prediction, positive mapping, and histogram packing using binary indexing vectors (positive packing) followed by a range coder. In order to achieve the optimal trade-off between residual errors and side information, an algorithm is proposed to determine adaptively the number of selected PCs and quantization parameters. Numerical experiments show that the proposed method outperforms the state-of-the-art methods (i.e. linear prediction with constant coefficients (LP-CC) and linear prediction with optimal granule ordering (LP-OGO)) by 1.77% in terms of compression ratio.