Analysis of approaches to hyperspectral data compression based on orthogonal transforms

Hyperspectral data are characterized by sufficient information content and large volume they occupy. This leads to necessity to compress these data with preserving high quality of images. Among existing lossy compression methods, approaches that employ multidimensional orthogonal transforms provide wide opportunities to satisfy a set of contradictory requirements to compression. However, due to specific properties of hyperspectral data, one can run into phenomena not mentioned and not studied well in the literature on processing hyperspectral images. This paper briefly analyzes possible approaches to compression based on wavelet and discrete cosine transforms and presents some recommendations for selecting methods able to control introduced distortions.