Abstract. We suggest a novel approach to the Color Constancy Problem for multispectral imagery. Our approach is based on a dichromatic illumination model and lters out all spectral information which possibly stems from the illumination rather than from the re ectance of a given surface. Instead of recovering the re ectance signal, the suggested mapping produces a new only surface re ectance-dependent descriptor which is invariant against varying illumination. Sole input is the relative direct to di use illumination spectrum, no assumptions about the possible re ectance spectra are made. The mapping is a purely pixel based, fast, one-pass matrix operation and can preprocess multispectral images in order to segment them into regions of homogeneous re ectance, unperturbed by varying illumination conditions.