On the ability of colour fidelity metrics to predict perceived colour difference

Cauwerts, Coralie [UCL] Jost, Sophie [ENTPE]

The Colour Rendering Index (CRI) defined by CIE is widely adopted and used to assess the ability of light sources to render colours. But over the past years, there has been increasing evidence of its limitations. Two technical committees have been set up by CIE, one dealing with colour fidelity of light sources (TC1-90) and the other with new methods for evaluating colour quality excluding fidelity (TC 1-91). TC1-90 will soon publish a Colour Fidelity Index (Rf) to predict the perceived colour difference between samples illuminated under two light sources (a test and a reference). This study was designed to evaluate the new fidelity index proposed by the CIE (Rf) through a psychovisual experiment. Performance of colour difference formulae and impact of colour matching functions were also investigated. The conducted experiment is based on the grey scale method to assess the colour difference of identical samples placed in two side-by-side booths. Twelve young observers (23-28 years old) with normal colour vision evaluated colour difference of 16 colour samples (+ 4 repetitions) under 8 pairs of light sources at two correlated colour temperature (3000K and 4000K). Eight colored samples were selected from the HL8 set developed for TC1-90 and the other eight were chosen from the Japanese Color Research Institute (JCRI) set. The light sources were LEDs and fluorescent light sources with different Rf values ranging from 60 to 90. The preliminary results suggest that colour difference calculated in CIECAM02-UCS colour space can predict the perceived colour differences and Rf is a rather good indicator of fidelity. CIE 2006 10° CMFs provides better visual match of the grey samples and showed better performance than the CIE 1931 colorimetric standard observer to describe visual results.