Hyperspectral images as function-valued mappings : their self-similarity and a class of fractal transforms

A formulation of hyperspectral images as function-valued mappings is introduced, along with a set of simple models of affine self-similarity for digital hyperspectral images. As in the case of greyscale images, these models examine how well vector-valued image subblocks are approximated by other subblocks, as measured by the distribution of approximation errors. This set of models includes both same-scale and cross-scale modes of approximation, the latter of which provides the basis of a method of fractal transforms over hyperspectral images