A Practical Non-Linear Parameterization of the BRDF Manifold - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Reports (Research Report) Year : 2017

A Practical Non-Linear Parameterization of the BRDF Manifold

Une parametrisation non linéaire mais versatile du manifolde des BRDFs


Real-world reflectance data can be used to improve the realism of synthesized images, albeit with many challenges: memory footprints can be large, profiles are limited to a finite (usually small) set of materials and rendering with measured data can be costly. Since the observation space (number of reflectance measurements) is usually much larger than the underlying space of real-world reflectance profiles, a typical optimisation strategy identifies principal components in the data to directly render from compressed representations of the measurements. We directly learn an underlying low-dimensional non-linear reflectance manifold amenable to rapid exploration and rendering of the space of real-world materials. We show that interpolated materials can be expressed as linear combinations of the measured data, despite lying on a non-linear manifold. This allows us to efficiently interpolate, extrapolate and render directly from the manifold. We apply a Gaussian process latent variable model to represent the reflectance manifold, demonstrating its utility in the context of high-performance and realistic rendering with materials that are interpolations of acquired BRDFs (from the popular MERL dataset).


Fichier principal
Vignette du fichier
RR-9069-1.pdf (8.87 Mo) Télécharger le fichier
Vignette du fichier
screenshot-area-2017-05-17-135827.jpg (23.88 Ko) Télécharger le fichier
Format : Figure, Image
Origin : Files produced by the author(s)

Dates and versions

hal-01523874 , version 1 (17-05-2017)


  • HAL Id : hal-01523874 , version 1


Cyril Soler, Kartic Subr, Derek Nowrouzezahrai. A Practical Non-Linear Parameterization of the BRDF Manifold. [Research Report] RR-9069, INRIA. 2017, pp.20. ⟨hal-01523874⟩
563 View
424 Download


Gmail Facebook Twitter LinkedIn More