CRI2012, a colour fidelity index for white light sources

Traditionally, the CIE colour rendering index has been used to described the colour rendition of white light sources. However, almost immediately after its standardization in 1974 there has been a growing body of literature and anecdotal evidence showing that it fails to accurately predict the colour rendition of narrow-band or spiked spectra. This has increasingly become a problem with the advent of Solid-State Lighting, as the inability to correctly assess the colour rendition of these light sources might hinder the acceptance of this new lighting technology. The failure has been shown to occur on two different fronts: 1) colour rendering or colour fidelity – an objective measure comparing the colour appearance under the test source with that under a reference illuminant; and 2) colour quality – referring to subjective aspects of colour rendition, such as preference, naturalness, vividness, harmony, … This paper focuses on the first aspect of colour rendition, i.e. colour fidelity. It summarizes the development of a new colour fidelity index, called the CRI2012, which improves upon some of the shortcomings of the CIE CRI calculation, like its outdated chromatic adaptation transform and perceptually non-uniform colour space, by updating them to the state of the art in colour appearance modeling. A new colour difference averaging and rescaling function is also proposed. But, most importantly, the CRI2012 replaces the Munsell test samples with two new sample sets: one spectrally uniform set, the HL17 set, to calculate a game-free general Ra,2012 index and another, the Real210, to provide hue specific information. The rationale behind the choice of the new and separate sample set for the general index is that uniformity in colour space is not a sufficient condition to ensure light source spectra cannot be gamed. What is needed is uniform sampling in wavelength space. At the moment the game-free set is still a mathematical set, but there is being actively worked on replacing it with a larger, spectrally uniform, set composed of real reflectance functions that adequately samples both wavelength and colour space, thereby replacing the need for the Real210 set. In addition, the possibility of updating the reference illuminant calculation is also being investigated. To avoid the discontinuity at 5000K – when the reference illuminant switches type – one might calculate the reference illuminant from, let’s say 4500K to 5500K, with a weighted mean of a Planckian radiator and a daylight phase, whereby the weighting is a smooth (non)-linearly increasing function of CCT.status: publishe