Context-based predictive lossless coding for hyperspectral images

A cluster-based lossless compression algorithm for hyperspectral images is presented. Clustering is carried out on the original data according to the vectors spectra, and it is used to set up multiple contexts for predictive lossless coding. Low-order prediction is performed using adaptive Linear Least Squares (LLS) estimation which exploits the additional information provided by clustering. Prediction errors are then entropy-coded using an adaptive arithmetic coder also driven by data clusters. The proposed scheme is used to losslessly code a set of AVIRIS hyperspectral images. Comparisons with the JPE-GLS, JPEG-2000 and the clustered DPCM coding algorithms are given.