Beyond Angular Error: Rethinking Metric for White Balance and Its Effect on Image Color Appearance

ABSTRACT White balance, which adjusts image colors based on the estimated illuminant color, has attracted significant attention in both academia and industry. Considerable efforts have been devoted to developing methods that more accurately estimate the illuminant color, typically aiming to minimize the angular error between the ground‐truth and estimated illuminants. The results of this study suggest that focusing solely on reducing angular error may not be optimal, given the practical variations in ground‐truth illuminants and the mismatch between angular error magnitude and the overall perceived color appearance of the processed image. Instead, minimizing the chromaticity shift between the estimated and ground‐truth illuminants perpendicular to the blackbody locus appears to be a more effective alternative to angular error.