Using Hopfield networks in the nominal color coding of classified images

Nominal color coding is widely used by the image processing community to represent the output of a classification-segmentation process. In order to automate color-class association we propose a suitable description scheme for both the color set to be used and the image to be coded. Using this description we then define a suitable energy function for Hopfield's neural networks. The objective is to associate more "conspicuous" colors with less "visible" classes, assigning highly contrasting colors to classes with a high "adjacency".