Dual-Direction Prediction Vector Quantization for Lossless Compression of LASIS Data

Abstract

Large Aperture Static Imaging Spectrometer(LASIS) is a new kind ofinterferometer spectrometer with the advantages of high throughputand large field of view. The LASIS data contains both spatial andspectral information in each frame which indicate the location shifting and modulatedoptical signal along Optical Path Difference(OPD). Based on these characteristics,we propose a lossless data compression method named Dual-directionPrediction Vector Quantization(DPVQ). With a dual-directionprediction on both spatial and spectral direction, redundancy inLASIS data is largely removed by minimizing the prediction residuein DPVQ. Then a fast vector quantization(VQ) avoiding codebooksplitting process is applied after prediction. Considering timeefficiency, the prediction and VQ in DPVQ are optimized to reducethe calculations, so that optimized prediction saves 60\% runningtime and fast VQ saves about 25\% running time with a similarquantization quality compared with classical generalized Lloydalgorithm(GLA). Experimental results show that DPVQ can achieve amaximal Compression Ratio(CR) at about 3.4, which outperforms manyexisting lossless compression algorithms.