Mean observer metamerism and the selection of display primaries

Abstract

Observer metamerism is an unavoidable fundamental limitation of colorimetry. While standard observers might provide accurate estimates of population-mean color matching functions, they can never predict accurate matches for all individual observers. This limitation is minimized in applications where the spectral difference (or potential for metamerism) is small. This is typically when original and reproduction both have slowly varying spectral power distributions. In the world of displays, that would be represented by broad-band primaries. As display manufacturers attempt to design displays with wider color gamuts and greater luminance contrast, narrow-band primaries (in some cases monochromatic laser primaries) are becoming more common. Such displays enhance the potential for significant differences in color perception and matching across individual observers. The CIE has recently published a method for computing color matching functions for mean observers of various ages and for various field sizes. This paper examines the interactions between these various color matching functions and display primaries. It is shown that the magnitude of mean-level observer metamerism can be significant (on the order of 4 CIELAB units) for broad-band primaries and a factor of two or more larger (on the order of10 CIELAB units) for narrow-band primaries. The potential disagreement between individuals is even larger and the discrepancies are particularly noticeable for large fields and nearwhite colors (as in trying to match the white point of two metameric projection displays or a display and metameric surround). Copyright 2007 Society for Imaging Science and Technology.