Lossless compression of hyperspectral images based on 3D context prediction

Abstract

Prediction algorithms play an important role in lossless compression of hyperspectral images. However, conventional lossless compression algorithms based on prediction are usually inefficient in exploiting correlation in hyperspectral images. In this paper, a new algorithm for lossless compression of hyperspectral images based on 3D context prediction is proposed. The proposed algorithm consists of three parts to exploit the high spectral correlation. Firstly, the LOCO-I prediction model similarity is chosen to set up 3D context prediction. Then a linear prediction algorithm is applied on the residual image after the 3D context prediction. Finally, the residual image of linear prediction is coded by the arithmetic coding. The performance of the proposed algorithm has been evaluated on AVIRIS hyperspectral images. The experimental results show that with a compression ratio (CR) up to 3.01, the proposed method obtains a better compression performance with comparison of partitioning DPCM, SSOLP, JPEG-LS, 3D-SPECK and 3D-SPIHT. The algorithm is of low complexity and can be implemented by FPGA or DSP for on-board implementation.