Real-time hyperspectral image cube compression combining adaptive classification and partial transform coding

In using partial transform for the coding of hyperspectral image, it must consider spectral decorrelation between image components, issues that a combined adaptive classification and partial transform algorithm for hyperspectral image compression is presented in this paper. Our method uses a linear prediction based on adaptive classification to decorrelate the spectrum redundancy and a 2D integer reversible DCT-based scheme as spatial compression engine. The classification includes band ordering, band regrouping and reference frame selection. Experimental results on AVIRIS data indicate that the proposed approach is a novel low complexity lossy hyperspectral image compression scheme and exhibits performance better than other comparative methods in quality and fidelity. It can be used for hyperspectral image compression in the embedded processor of the platform on satellite or aircraft