Context Modeling in Problems of Compressing Hyperspectral Remote Sensing Data

The article is devoted to developing a compression method using context modeling of a sequence of bits and wavelet transform, which make it possible to take into account the specifics and properties of the initial hyperspectral remote sensing data. Two algorithms for compressing hyperspectral data (lossy and lossless) based on wavelet transform are proposed, the distinguishing features of which are reduction in the required memory size, acceleration of the search for significant wavelet coefficients using a pyramid with approximating coefficients, and an increase in the compression coefficient. Recommendations for applying these algorithms are formulated. A distinctive feature of the hyperspectral data compression method is the ability to control the compression coefficient owing to parametric adjustment of the algorithms, application of context modeling and adaptation to the type of initial data (classical cube or Fourier interferogram). The efficiency of the technique has been experimentally confirmed using examples of compression of classical data and real Fourier interferograms with compression ratios of 4.1 and 2.4, corresponding to the level of the best global results, as well as analytically with data distortion in a compressed stream.