Testing colour-difference formulas from LMS colour spaces inspired in CIELAB

Abstract

The CIELAB was recommended in 1976 as an approximately uniform colour space and colour-difference formula ( $\Delta E_{\mathrm{ab}}^{*}$ ) based on CIE XYZ tristimulus values. In this paper, first CIELAB is improved by keeping its structure and optimizing its four parameters in XYZ, LMS and X_FY_FZ_F spaces, resulting in modifications named MLAB(XYZ), MLAB(LMS) and MLAB(X_FY_FZ_F), respectively. Then, CIELAB and these three modifications are evaluated in XYZ, LMS and X_FY_FZ_F colour spaces, respectively. From the combined and corrected (COM-corrected) visual dataset employed after the development of CIEDE2000 (Journal of the Optical Society of America A 25, 1828–1834, 2008), using the standardized residual sum of squares (CIE 217:2016), we found that CIELAB and its modifications perform similarly in XYZ and X_FY_FZ_F spaces and perform worse in LMS colour space. It was also found that the three modifications to CIELAB performed significantly better than the original CIELAB in predicting the COM-corrected dataset, but significantly worse than the CIEDE2000 colour-difference formula (ISO/CIE 11664-6:2022), CAM16-UCS (Color Research and Application 42, 703–718, 2017) and simple uniform colour space (Optics Express 32, 3100–3122, 2024) spaces. The MLAB(X_FY_FZ_F) space can therefore be tested as an approximately uniform colour space and colour-difference formula, following research proposed by CIE Technical Committee 1-98 (CIE 254:2024).