Projector radiometric compensation using a 2D spectroradiometer

Apr 20, 2026

Optics Express

Abstract

Projection mapping (PM) optically overlays computer-generated imagery onto realworld objects, enabling users to experience augmented reality without wearing any display devices. However, surface textures often cause color distortion, leading the displayed colors to deviate from the desired colors. To address this issue, we propose a projector radiometric compensation method that minimizes the color difference between a target image and the projected result using a 2D spectroradiometer (2DSR). In the proposed method, we model the color transformation between the projector and the 2DSR in a differentiable manner. Based on this formulation, we propose two optimization strategies for projector radiometric compensation: (i) minimizing the spectral error between the target appearance and the projected result, and (ii) minimizing the color difference measured in a differentiable color space designed to reflect human visual perception. Experiments with a physical prototype demonstrate that our method achieves more accurate projector radiometric compensation and better alignment with human color perception than conventional methods using an RGB camera.

Experimental setup, consisting of a projector (EPSON EB-FH52), a 2D spectroradiometer (TOPCON SR-5100), and the projection surface (paper with inkjet-printed patterns). The projector-to-surface distance was 0.8 m.
Experimental setup, consisting of a projector (EPSON EB-FH52), a 2D spectroradiometer (TOPCON SR-5100), and the projection surface (paper with inkjet-printed patterns). The projector-to-surface distance was 0.8 m.
Comparison of projection results across objective functions (ablation study). (a) sRGB-converted visualization images of 2DSR-captured projections, with ∆E76 and spectral-MAE heatmaps relative to the target spectral image. (b) Mean-spectra comparison within the marked rectangular region. Evaluates Ospe, Orgb, Olab and all their combinations.
Comparison of projection results across objective functions (ablation study). (a) sRGB-converted visualization images of 2DSR-captured projections, with ∆E76 and spectral-MAE heatmaps relative to the target spectral image. (b) Mean-spectra comparison within the marked rectangular region. Evaluates Ospe, Orgb, Olab and all their combinations.
Convergence behavior of the proposed iterative optimization, with evaluation metrics (spectral MAE, PSNR, ∆E76) plotted as a function of the iteration number.
Convergence behavior of the proposed iterative optimization, with evaluation metrics (spectral MAE, PSNR, ∆E76) plotted as a function of the iteration number.
Comparison of projection results across methods, contrasting the baselines (Yoshida et al., CompenNeSt w/SL) with the proposed method (Orgb+Olab, Ospe+Orgb+Olab). (a) visualization images and error heatmaps, (b) mean-spectra comparison. The proposed method most closely reproduces the target spectral shape.
Comparison of projection results across methods, contrasting the baselines (Yoshida et al., CompenNeSt w/SL) with the proposed method (Orgb+Olab, Ospe+Orgb+Olab). (a) visualization images and error heatmaps, (b) mean-spectra comparison. The proposed method most closely reproduces the target spectral shape.

Citation

Yoshiaki Maeda and Daisuke Iwai, “Projector radiometric compensation using a 2D spectroradiometer,” Optics Express 34(9), 15979–15993 (2026).

Projects XR Displays
Previous