Using Optimal Surface Re ectance Functions to Study

We present a computational deenition of the Optimal Color Stimuli surface, which corresponds to the limit of physically realizable surface spectral reeectance functions. We can use this surface to quantify and visualize color space properties, as we demonstrate for the CIE XYZ and L*a*b* spaces, as well as our own neurophysiologically-based NPP space. We then extend the technique to compute a representative set of surface spectral reeectance functions, all contained within the volume bounded by the OCS surface, and their corresponding representations in a color space. This set of spectra can be used to generate training sets for a backpropagation algorithm computing color space transforms, for example.