Spectrally guided color calibration method for the full imaging-to-display chain

Conventional color reproduction uses fragmented optimization across imaging and display stages, compromising consistency between real scenes and display output. To address this problem, we propose a unified, spectrally grounded framework. Hyperspectral measurements establish device-independent XYZ targets, a 3D-LUT generates display-driving RGB labels, and a fully convolutional model learns a direct RAW RGB-to-display mapping. ColorChecker validation yields a mean CAM16 response distance and CIEDE2000 color difference of 0.190 and 1.793, respectively, versus 2.444 and 12.007 for direct JPEG display. Under the manually established white-reference correspondence used in this study, tests on 10 displays, real-scene comparisons, and subjective evaluations consistently demonstrate markedly improved color fidelity.