A grid approach to optimizing color recipes.

We show that the accuracy of predicting color recipes for solid colors using the conventional Kubelka-Munk model can be improved by using grid-based empirical techniques.We first identify the regions in color space where such improvements would be most useful. These turn out to be the dark red, dark blue and the off-white regions. In these regions of color space, we create grids using several different methods: Delaunay triangulation and Adaptive meshing techniques with thresholds either on non-linearity or on fixed distances between grid points. This last method was shown to work best for this application.In order to match an arbitrary point in color space, based on the created grid, two different interpolation methods were tested: Linear optimization (where a linear relation between concentration space and color space is assumed within each grid cell) and Local K and S determination (where values of the Kubelka-Munk parameters are allowed to vary over color space). Our results show that grid methods using Local K and S determination lead to a significant improvement in accuracy as compared to conventional Kubelka-Munk methods.