Nomograph functions and image representation

An approach that combines three ideas: nomographs, the singular-value-decomposition (SVD), and histogram modification to image representation and compression, is presented. The basic idea is that most image representation techniques use multiplicative-separable function (e.g. SVD, Fourier), but these are not always optimal. A larger class of functions, called nomographic functions, is introduced. It is shown how this class of functions is related to separability and histogram modification. An example of an image estimator/representer based on these ideas is given. It is shown that it can slightly outperform the first component SVD