Newton transforms: new tools for cast and image processing

Abstract

From Newton filters, previously presented computer aided systems theory (CAST) tools whose basic ideas for design are inspired directly by the neurons of the retinal pathway, we define a new class of transforms, the Newton transforms, using what we call Newton matrices. The properties of these matrices, which are explained in this paper, make it possible to define this transformation on data fields whose inverse has the same structure as the direct transform. A parallel processing neuron-like structure for computing these calculations is designed, and an example of the use of this technique is on image processing and coding is also presented.