Sampling CIELAB color space with perceptual metrics

Abstract

Sampling a perceptually uniform or pseudo-uniform color space is required for applications from image processing to computational imaging. However, one can face two problems while trying to perform a uniform sampling of such space. First, the usual cubic grid is not perceptually uniform in most cases. Second, perceptual metrics are often not Euclidean. We propose to overcome these problems. We apply our solution on CIELAB color space to test its efficiency. We propose an algorithm to define a tabulated color space with regard to a non-Euclidean color difference formula, i.e. E 00 in CIELAB. The tabulated data are available at http://data.couleur.org/deltaE/. Later, we propose to combine this tabulated color space with an approximated 3D close packed hexagonal regular sampling of CIELAB. Evaluations of the transform and of the regular sampling are performed and compared with literature standards.