Accurate spotlight spectra estimation using neural methods and compact disc diffraction gratings
Year: 2026
Type: article
Abstract: Inverse rendering and appearance modeling increasingly rely on accurate spectral modeling of scene illumination to synthesize photo-realistic images. However, measuring an illuminant’s spectral power distribution (SPD) using a spectroradiometer or hyperspectral imaging systems is costly and typically confined to specialized laboratories. We introduce an inexpensive, imaging-based method to estimate precise spotlight illumination SPDs by exploiting wavelength dependent diffraction patterns produced by a commercially available compact disc (CD). Our method captures rainbow-like diffraction rings produced by shining an unknown spotlight on a CD using a consumer grade camera, and (optionally) augments it with a color chart image for reliable spectral reconstruction of the spotlight illumination without needing explicit geometric or color calibration. Comprehensive qualitative and quantitative experiments on both physically based simulations and real-world images show that diffraction cues result in consistent high frequency spectral details that are hard to infer from a color chart alone. We assess the generalizability of our approach by evaluating it using 49 distinct (simulated) camera models including Nikon, Canon, Sony, iPhone and Samsung Galaxy. Furthermore, we examine the robustness of our method under manual flexibility in the placement and orientation of the compact disc where both can be easily constrained through quick visual inspections. Lastly, we validate the practical impact of our reconstructed SPDs through spectral inverse rendering of diffractive and iridescent materials, showing improved visual fidelity and reduced error.
Source: Graphical Models
Authors: Parisha Joshi, Daljit Singh J. Dhillon
Institution: Clemson University
Language: English
Subfield: Atomic and Molecular Physics, and Optics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,180
OpenAlex ID: W7169849669
