Multi-spectral data compression using hybrid cluster coding

An adaptive technique achieving data compression of multi-spectral images through the use of clusters is studied. The algorithm consists of a modified multidimensional (multi-spectral) extension of a mode-seeking estimator due to Koontz et al. (1976) followed by weighted minimum distance classifier. This estimator is applied to 16 x 16 blocks of Landsat multi-spectral scanner (MSS) data across the four spectral bands to select the number and relative positions of the modes of the data. Excellent performance is demonstrated in terms of percent distortion of original and reconstructed data blocks.