Lossless Compression of Hyperspectral Images Using Interband Gradient Adjusted Prediction

Interband coding techniques are needed for effective compression of hyperspectral images, since high interband correlation cannot be exploited by intraband prediction. In this letter, an interband version of GAP (gradient adjusted prediction) is proposed by combining a linear prediction with a gradient adjusted prediction. The corresponding prediction function is chose by comparing the difference between the estimate of horizontal gradients and that of vertical gradients with a given threshold. After prediction, the difference is entropy-coded using an adaptive entropy coder. Experimental results on Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data show the proposed algorithm can exploit both interband and intraband statistical correlations, and achieve better compression performance compared with those existing classical algorithms. Moreover, low encoder complexity makes it suitable for on-board compression of hyperspectral images.