Hyperspectral imagery compression based on linear spectral mixture theory

It is a challenge to efficiently compress hyperspectral imagery with smaller compression error on the satellite platform. This paper presented a novel method for hyperspectral imagery compression that analyzed hyperspectral imagery based on linear spectral mixture analysis. Algorithm of vertex component analysis(VCA) is applied to extract endmembers of the imagery and some of them selected according to channel capacity for mixed pixel analysis. The fractional abundances of the selected endmembers at all the pixels are then computed using linear spectral unmixing method. The selected endmembers and their fractional abundances are encoded using the arithmetic of JPEG 2000 lossless compression. Experiments on the hyperspectral imagery of AVIRIS indicate: at an 80:1 compression ratio, the maximum relative error of the presented method does not exceed 2.7%, and the maximum spectral angle cosine error is less than 0.00023, the compression performance better than that of the existing algorithm. Furthermore, the method can suppress random noise in the original hyperspectral imagery.