30 January 2003 A Compression algorithm for hyperspectral imagery using the linear mixing model and wavelet transform
Author Affiliations +
Abstract
A scheme for lossy hyperspectral data cube compression, using a linear mixing model approach and wavelet transform, is presented. The data is first compressed in the spectral dimension by using the linear mixing model approximation to reduce the number of dimensions needed to represent the data. The reduced data is then compressed along the spatial dimensions using a wavelet transform. Five hyperspectral data cubes have been tested using the algorithm. Compression ratios of up to 1000:1 are achieved with peak signal-to-noise (PSNR) ratios of over 40 dB. For all test cases, we were able to achieve ratios of over 200:1 with PSNR exceeding 46 dB. The ultra-high compression ratio with low distortion is an improvement over other results reported in the literature. In addition, the reconstructed spectra from the highly compressed file are shown to preserve the overall shape of the original spectra. However, in some cases the curves are slightly offset in some spectral regions from the original.
© (2003) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Wei Chen, Wei Chen, David Gillis, David Gillis, Jeffrey H. Bowles, Jeffrey H. Bowles, Curtiss O. Davis, Curtiss O. Davis, } "A Compression algorithm for hyperspectral imagery using the linear mixing model and wavelet transform", Proc. SPIE 4793, Mathematics of Data/Image Coding, Compression, and Encryption V, with Applications, (30 January 2003); doi: 10.1117/12.463646; https://doi.org/10.1117/12.463646
PROCEEDINGS
12 PAGES


SHARE
Back to Top