How color difference formulas depend on reference pairs in the underlying constant stimuli experiment

Area of Science:

• Color Science
• Metrology
• Psychophysics

Background:

• The CIEDE2000 and CIE94 color difference equations are industry standards, largely based on the RIT-Dupont dataset derived from psychophysical tests.
• These established equations rely on a specific method of constant stimuli, where observers compare sample color differences to reference pairs.

Purpose of the Study:

• To investigate how variations in reference pairs within visual experiments affect the performance of color difference equations.
• To determine if current models, like CIEDE2000, adequately represent color perception with metallic samples and alternative reference pairs.

Main Methods:

• Utilized metallic paint samples (red and yellow-green color centers) in psychophysical tests.
• Employed the method of constant stimuli, systematically varying reference pairs compared to the original RIT-Dupont dataset's references.
• Analyzed the reproducibility and required model adjustments for different reference pairs.

Main Results:

• Color difference equation outcomes significantly change when reference pairs deviate from those used in the RIT-Dupont dataset.
• Reproducibility varied across three tested reference pairs, with the yellow-green center requiring additional model terms beyond CIEDE2000.
• Observed widened tolerance for lightness differences (1.3-1.6x) and a 30° rotation of the tolerance ellipsoid for the red center, potentially due to metallic texture.

Conclusions:

• The choice of reference pairs in psychophysical testing critically impacts color difference equation results, challenging the universality of current models.
• Observer variability in mentally equating color differences and sample properties (like metallic texture) necessitate refined or new color difference models.
• Existing color difference equations may require modifications, particularly for non-ideal samples like metallic paints, to ensure accurate perceptual representation.