A Modular Approach for Leveraging BRDF Monotonicity in Computing Shading Normals from Multi-Light Image Collections

Multi-Light Image Collections (MLICs) are stacks of photos acquired with a fixed viewpoint and varying surface illumination. They are typically used to provide rich information to characterize imaged objects. In particular, many simple and sophisticated Photometric Stereo solutions exist to produce detailed per-pixel normal maps that geometrically characterize the imaged model up to the finest details. These maps are commonly employed for per-pixel BRDF fitting to generate easy-to-use general shape and material representations for visual exploration. However, discrepancies between the chosen BRDF space, the complete optical behavior of complex objects, and the presence of non-local lighting effects in measurements may lead to suboptimal visual outcomes even with the most accurate geometric normal recovery. This article introduces a modular component designed to convert the geometric normals into well-behaved shading normals, under the common and general assumption that the reflectance under local illumination must be a monotonic function of the angle between the shading normal and the bisector of lighting and viewing directions. Since it does not require the coupling of shape and material estimation, the module allows seamless integration into existing reconstruction pipelines, supporting the mixing and matching of Photometric Stereo methods, BRDF models, and BRDF fitters. We demonstrate our approach’s efficacy as a versatile solution for enhancing the fidelity of rendered images derived from MLICs through synthetic data, laboratory experiments, and real-world painting measurements.

@Article{PZG25,
  author       = {Pintus, R. and Zorcolo, A. and Gobbetti, E.},
  title        = {A Modular Approach for Leveraging BRDF Monotonicity in Computing Shading Normals from Multi-Light Image Collections},
  journal      = {ACM Journal on Computing and Cultural Heritage (JOCCH)},
  year         = {2025},
  publisher    = {ACM Press},
  note         = {To appear},
  keywords     = {visual computing,cultural heritage},
  url          = {https://publications.crs4.it/pubdocs/2025/PZG25},
}