Compression of Multispectral Images

Compression of multispectral images (obtained from sensors that sample in both the spatial and the spectral domains) is important for reducing the transmission and storage requirements of such data. This article describes versatile software developed to simulate a family of compression algorithms. Two algorithms are selected as the most suitable for implementation. The first is a moderately high complexity algorithm consisting of the Karhunen–Loève transform (KLT) in the spectral dimension, the discrete cosine transform (DCT) for spatial decorrelation of the resulting bands, and the DCT on the residual. The other is a mediumcomplexity algorithm that uses predictive coding for spectral decorrelation and the DCT for spatial decorrelation. Performance results are given for these algorithms. A low-complexity algorithm is also discussed.