Quantifying lighting consistency in Impressionist painting
is article develops a computational approach to pictorial lighting in Impressionist painting.While Impressionism is often associated with the observation of natural light, its paintings also involve constructed, studio-based, and multi-source lighting effects.We ask whether such effects can be studied systematically across a large corpus without treating computational estimates as reconstructions of original illumination.Using 1,394 digitized Impressionist paintings by ten artists, we apply a computer-vision pipeline that estimates lighting direction and luminance distribution from depicted figures and synthesized chrome-ball renderings.We introduce "lighting consistency" as an interpretively bounded proxy for the coherence of illumination within a painting.e analysis compares indoor and outdoor scenes, artists, and time periods through measures of light elevation, angular separation, luminance concentration, angular spread, inlier fraction, and mean angular error.Results show that outdoor scenes exhibit higher estimated light elevation, while indoor compositions display more localized and directionally varied illumination.Most paintings also show evidence of multiple or spatially distributed light components in the estimated environments, suggesting that constructed lighting was a broadly shared feature of the figure paintings examined in this dataset.A temporal decline in angular error is consistent with a trend toward greater concentration of estimated light directions over time, although this pattern warrants further investigation.By translating an art-historical concept into measurable form while preserving interpretive caution, the study demonstrates how computer vision can support large-scale, humanistically grounded analysis of artistic practice and visual style.
Authors
- Sitao Zhang (ORCID: https://orcid.org/0009-0005-7336-6753)
- Kathryn Brown (ORCID: https://orcid.org/0000-0001-5710-1708)
- Anushcka Joshi (ORCID: https://orcid.org/0009-0009-7805-9153)
- Lizhen Zhu (ORCID: https://orcid.org/0000-0002-4538-5640)
- James Z. Wang
Institutions
- Pennsylvania State University (US)
- Loughborough University (GB)
Publication Details
- Journal
- Computational Humanities Research
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1017/chr.2026.10052
- Primary Topic
- Computer Graphics and Visualization Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00