Evaluating GOG, PLCC and PLVC and their perceptual limits in optical see-through augmentation reality

Accurate colour prediction for head-mounted displays (HMDs) is limited by the interaction between display output, near-eye optics, interocular viewing paths, background transmission and human perception. This study evaluates three established point-colour characterisation models—gain-offset-gamma (GOG), polynomial linear-channel combination (PLCC) and polynomial linear-vector combination (PLVC)—across five conventional planar displays and three representative HMDs − HTC VIVE Pro, Meta Quest 3 and Microsoft HoloLens 2. Spectroradiometric measurements were conducted under controlled conditions using 93 colour samples per device, including single-channel ramps, mixed-channel samples and ColorChecker patches. Conventional planar displays showed low colour-difference errors and relatively stable additive behaviour. In contrast, HMDs produced larger residual errors, interocular differences, channel-dependent chromatic shifts and architecture-dependent model performance. The HoloLens 2 showed the most complex behaviour, including elevated black levels, left–right gamut displacement, measured-gamut contraction relative to its nominal gamut and non-linear channel trajectories, consistent with the influence of laser-scanning and diffractive-waveguide optics. The paper further proposes a three-level distinction between nominal display gamut, through-optics measurement gamut and perceptual gamut for optical see-through HMDs. Observer-based perceptual experiments, including luminance matching and colour matching under HoloLens 2 viewing conditions, showed that perceived matches were affected by background condition, colour region and the through-optics appearance of physical references. Observer-selected colours did not consistently coincide with either nominal device targets or instrumentally measured references. These results indicate that low instrumental ΔE 00 does not necessarily ensure perceptual appearance matching in optical see-through AR. GOG, PLCC and PLVC remain useful baseline models, but reliable colour prediction for near-eye displays requires per-eye optical consideration, black-level handling, architecture-dependent interpretation, background interaction analysis and observer-based validation.