In this study, the observer variability in a 10° achromatic color matching experiment using systematically different primaries under immersive, perceptually neutral (4673 K) and stable adapted viewing conditions has been investigated for 4 different sets of CIE color matching functions (CMFs). The matching reference was illuminated by a spectrally broadband light source at 70 cd/m² and 4613 K. Fifty-four observers varying in age, gender, ethnicity and eye color participated. The results indicate that certain wavelength regions are more sensitive to generate inter-observer variability, such as when a short wavelength primary (404 nm) was applied. Furthermore, of the primary sets investigated, that with peak-wavelengths (636-521-447 nm) close to the 'typical' RGBprimaries widely known and used in industry had the smallest inter-observer variability. In addition to the significant effect of primary set and CMF set, a significant effect of age on matching variability is also found, whereby the change of the blue primary resulted in a substantial difference in dispersion between younger and older observers, likely caused by age-induced yellowing of the lens. Furthermore, different primary sets tend to have a different impact on intra-observer variability than on inter-observer variability. Finally, by comparing the observer variability in the current work with previous studies, the CIE standard deviate observer is found to substantially underestimate the inter-observer variability, especially for the primary sets containing a 404 nm LED. Our observers also tend to have similar to smaller inter-variability than those reported in literature, possibly due to methodological differences.

No abstract

No abstract

Accurate colorimetry starts with accurate color matching functions (CMFs). Due to changes in the macular pigment and cone pigment optical densities at different retinal locations, different CMFs are required for different stimulus field sizes. To characterize the accuracy of the CIE recommendation for the use of 2° and 10° standard CMFs and the field‐size dependent CIEPO06 model, in this study, a series of achromatic matching experiments were performed with 2°, 4°, 6°, 8°, and 10° bipartite fields using spectrally narrowband primaries. Using the CIEPO06 model, optimal field sizes were estimated that minimize the chromaticity differences between the spectrally narrowband observer matches and the spectrally broadband achromatic target. It was found that the optimal field size estimated using the CIEPO06 model is close to half the actual bipartite field size in most cases, except for the 2° field. The discrepancy between the 10° bipartite field in Stiles & Burch's experiments and our optimal field size (6.54°) was assumed to be due to different individual color comparison strategies. However, the results of experiments requiring specific observer gaze positions did not support this assumption and the reason for the discrepancy remains unclear. Confirming our earlier results, the primary set (636, 521, 447 nm) was again found to be largely insensitive to changes in CMFs and to provide the most accurate matches under various fields of view. The inter‐ and intra‐observer variability ellipses for 2° matches are larger than those for 10°, consistent with Brown's finding for color discrimination ellipses. The magnitude of the intraobserver variability was similar for all field sizes, except for 2° field size, where matching errors were larger for some primary sets.

Over time, much work has been carried out to ascertain the accuracy of the CIE standard color-matching functions, but no definitive answer has been given. Recent work indicates an undeniable discrepancy between visual and computed metamers calculated using the existing CIE (the International Commission on Illumination) standard observer CMFs, especially when matching with narrowband sources. With a spectrally tunable solid-state light source, a series of pilot matching experiments have been done using primaries with different peak wavelengths. The results indicate which regions in wavelength space are most sensitive to generating matching inaccuracies for a given CMF set and which primary combinations have the most stable matching performance.