Block-based KLT compression for multispectral images

Abstract

An efficient lossy compression algorithm for multispectral images based on block Karhunen–Loève transform (KLT) is proposed. First, a two-dimensional discrete wavelet transform is performed on each band of multispectral images to remove the spatial correlation. Subsequently, each band is partitioned into non-overlapping blocks of the same size, and the transform coefficients of each block in the wavelet domain are treated as a single object. Blocks that are co-located in the spectral orientation are affected by an adaptive Karhunen–Loève transform to remove their spectral correlation. Finally, embedded block coding with optimized truncation is performed on all principal components to generate the final bit-stream. Experimental results show that the proposed algorithm, based on block KLT, outperforms the algorithm based on global KLT, without significant increase of complexity.