Multispectral image compression by cluster-adaptive subspace representation

Multispectral imaging has attracted much interest in color science area, for its ability in providing much more spectral information than 3-channel color images. Due to the huge data volume, it is necessary to compress multispectral images for efficient transmission. This paper proposes a framework for spectral compression of multispectral image by using cluster-adaptive subspaces representation. In the framework, multispectral image is initially segmented by hierarchical analysis of the transform coefficients in the global subspace, and then ambiguous pixels are identified and classified into proper clusters based on linear discriminant analysis. The dimensionality of each adaptive subspace is determined by specified reconstruction error level, followed by further cluster splitting if necessary. The efficiency of the proposed method is verified by experiments on real multispectral images.