Boundary-Continuous Spectral Demoiréing for Pixel-Level Luminance Measurement of Display Panels

Camera-based pixel-level luminance measurements are susceptible to moiré artifacts arising from aliasing between the display-pixel array and the camera sampling grid. Although frequency-domain notch filtering can attenuate periodic interference, the finite extent of a display panel, particularly its straight edges, rounded corners, and cutouts, introduces spectral leakage and boundary-related ringing. This paper proposes boundary-continuous spectral demoiréing (BCSD), which constructs a peripheral image extension before notch filtering. A uniform W192 reference is used to calibrate a configuration-specific spectral mask. Guided by this mask, alternating frequency-domain and spatial-domain constraints generate an extended image, which is subsequently notch-filtered and cropped to the original panel region. The method is evaluated against conventional notch filtering and representative learning-based approaches using OLED and LCD panel measurements acquired under three camera mappings and three color channels. The evaluation comprises 90 image–configuration records, including 18 W192 records also used for offline calibration. BCSD achieves a mean peak signal-to-noise ratio (PSNR) of 36.602 dB and a mean structural similarity index (SSIM) of 0.9334, exceeding conventional notch filtering by 1.524 dB and 0.0107, respectively. These results establish improvements in restoration relative to digital reference images under the evaluated acquisition conditions.