Decorrelate hyperspectral images using spectral correlation

This paper proposes a new algorithm for lossless compression of hyperspectral images. In our work we found hyperspectral data have unique characteristic based on spectral context and adjacent pixel spectral vectors (curves) highly correlate with each other. Pearson correlation coefficient is an effective measure of spectral similarity between spectral curves to detect horizontal and vertical spectral edge. Thus, spectral correlation is used to prediction in spectral direction for decorrelation of lossless compression of hyperspectral images. Experiments show the proposed algorithm is effective, and it's more important that it has much lower complexity than other algorithms.