Awesome-BRDF

A curated reading list on BRDF representations, measured reflectance, neural materials, and BRDF estimation in computer vision and computer graphics.

Last literature review: 2026-10-11. Years refer to the linked published version when available; preprints, technical reports, and course notes are labelled explicitly. This is a selection of foundational and representative work, not an exhaustive bibliography. See review notes for scope and corrections.

Table of contents

• BRDF Representation
  • Notation and physical properties
  • Analytic and microfacet models
  • Layered materials and multiple scattering
  • Measured representations, fitting, and sampling
  • Neural BRDF representations
  • Fresnel Term
  • BRDF formulas
• SVBRDF capture and inverse rendering
• Datasets and implementations
• Surveys and further reading
• Contributing

BRDF Representation

Notation and physical properties

The BRDF relates incident irradiance to outgoing radiance at a surface point. Both unit directions point away from the surface: $\omega_i$ toward the light and $\omega_o$ toward the viewer. Let $\mathbf n$ be the surface normal, $\mu_i=\mathbf n\cdot\omega_i>0$, $\mu_o=\mathbf n\cdot\omega_o>0$, and $\mathbf h=(\omega_i+\omega_o)/|\omega_i+\omega_o|$. The equations below apply to reflection in this upper hemisphere.

$$f_r(\omega_i,\omega_o)=\frac{\mathrm d L_o(\omega_o)}{\mathrm d E_i(\omega_i)}, \qquad L_o(\omega_o)=\int_{\Omega^+}f_r(\omega_i,\omega_o)L_i(\omega_i)\mu_i\,\mathrm d\omega_i.$$

For passive, reciprocal materials without wavelength conversion:

$$f_r(\omega_i,\omega_o)\geq 0, \qquad f_r(\omega_i,\omega_o)=f_r(\omega_o,\omega_i), \qquad \int_{\Omega^+} f_r(\omega_i,\omega_o)\mu_o\,\mathrm d\omega_o\leq 1.$$

The integral bound holds for every incident direction and wavelength; in an RGB approximation it is checked per channel. A BRDF value can exceed one: the bound is on the cosine-weighted integral. Energy preservation for a lossless reflector means equality, whereas energy conservation permits absorption. A physically based model or a small training loss alone does not establish either reciprocity or conservation for every parameter setting.

An SVBRDF adds surface position, $f_r(\mathbf x,\omega_i,\omega_o)$. A neural predictor of analytic BRDF parameters and a network that directly predicts BRDF values are different representations. BSDF papers below also cover transmission; their reflection components are relevant here. BTFs include spatial appearance effects such as occlusion and are not interchangeable with local SVBRDFs.

Analytic and microfacet models

Year	Publication	Model / authors	Paper	Representation / caveat
1760	Book	Lambert	Photometria sive de mensura et gradibus luminis, colorum et umbrae	Historical basis of Lambertian reflection; modern BRDF is albedo divided by pi.
1967	JOSA	Torrance–Sparrow	Theory for Off-Specular Reflection From Roughened Surfaces	Geometric microfacet reflection with masking and shadowing.
1975	CACM	Phong	Illumination for computer generated pictures	Empirical illumination model; distinguish its original shading equation from normalized BRDF variants.
1977	SIGGRAPH	Blinn–Phong	Models of light reflection for computer synthesized pictures	Half-vector specular lobe; normalization depends on the variant.
1982	TOG	Cook–Torrance	A Reflectance Model for Computer Graphics	Microfacet specular model; the linked journal version is 1982.
1991	SIGGRAPH	He et al.	A comprehensive physical model for light reflection	Physical rough-surface reflection including wave effects.
1992	SIGGRAPH	Ward	Measuring and modeling anisotropic reflection	Compact anisotropic lobe; the original model is not uniformly energy conserving.
1994	SIGGRAPH	Oren–Nayar	Generalization of Lambert's reflectance model	Rough diffuse reflection; distinguish full and qualitative approximations.
1994	CGF / Eurographics	Schlick	An Inexpensive BRDF Model for Physically-based Rendering	Efficient reflectance model and the widely used Fresnel approximation.
1997	SIGGRAPH	Lafortune et al.	Non-Linear Approximation of Reflectance Functions	Generalized cosine lobes; fitted coefficients need physical constraints.
1999	CGF / Eurographics	Neumann et al.	Compact Metallic Reflectance Models	Physically plausible modifications of empirical specular models.
2000	JGT	Ashikhmin–Shirley	An anisotropic Phong BRDF model	Reciprocal, energy-conserving anisotropic model.
2006	TOG	Edwards et al.	The halfway vector disk for BRDF modeling	Halfway-vector disk representation.
2007	Technical report	Ashikhmin–Premoze	Distribution-based BRDFs (archive lookup)	Distribution-driven reflection; original author page returns 404. Archive lookup is provided; full text was not recovered in this review.
2007	EGSR	Walter et al. / GGX	Microfacet Models for Refraction through Rough Surfaces	Introduces GGX in this BSDF formulation; reflection and transmission. Single scattering can lose energy on rough surfaces.
2010	ACM SIGGRAPH Computer Graphics	Kurt et al.	An Anisotropic BRDF Model for Fitting and Monte Carlo Rendering	Fittable anisotropic model with importance sampling; a journal article, not a SIGGRAPH technical paper.
2012	CGF / Eurographics	Bagher et al.	Accurate fitting of measured reflectances using a Shifted Gamma micro-facet distribution	Shifted Gamma NDF fitted to measured materials.
2012	TOG	Löw et al.	BRDF models for accurate and efficient rendering of glossy surfaces	Models for measured glossy reflectance.
2012	SIGGRAPH course notes	Burley / Disney	Physically-Based Shading at Disney	Artist-oriented principled BRDF; physical guarantees depend on lobe composition and parameters.
2014	TOG / SIGGRAPH	Brady et al. / genBRDF	genBRDF: discovering new analytic BRDFs with genetic programming	Automatic discovery of analytic expressions.
2014	TOG / SIGGRAPH	Jakob et al.	Discrete stochastic microfacet models	Discrete microstructure for glints.
2014	JCGT	Heitz	Understanding the Masking-Shadowing Function in Microfacet-Based BRDFs	Derivation and normalization of masking-shadowing functions.
2025	JCGT	Portsmouth, Kutz, Hill / EON	EON: A practical energy-preserving rough diffuse BRDF	Analytical energy compensation for rough diffuse reflection; 2024 preprint, 2025 publication, revised 2026.

Layered materials and multiple scattering

Year	Publication	Authors	Paper	Main idea
2001	CGF / Eurographics	Ershov et al.	Rendering Pearlescent Appearance Based On Paint-Composition Modelling	Composition-based pearlescent paint appearance.
2007	GRAPHITE	Weidlich, Wilkie	Arbitrarily Layered Micro-Facet Surfaces	Multiple microfacet layers; corrects the previous 2008 entry.
2014	TOG / SIGGRAPH	Jakob et al.	A Comprehensive Framework for Rendering Layered Materials	General layered transport with precomputed representations.
2016	TOG / SIGGRAPH	Heitz et al.	Multiple-Scattering Microfacet BSDFs with the Smith Model	Recovers the multiple-scattering component omitted by single-scattering Smith models.
2017	SIGGRAPH course slides	Kulla, Conty	Revisiting Physically Based Shading at Imageworks	Practical energy compensation and layered production shading.
2017	TOG / SIGGRAPH	Belcour, Barla	A Practical Extension to Microfacet Theory for the Modeling of Varying Iridescence	Thin-film interference and a spectral Fresnel extension.
2018	TOG / SIGGRAPH	Belcour	Efficient Rendering of Layered Materials using an Atomic Decomposition with Statistical Operators	Statistical operators for efficient layered BSDF approximation.
2018	TOG / SIGGRAPH	Zeltner, Jakob	The Layer Laboratory: A Calculus for Additive and Subtractive Composition of Anisotropic Surface Reflectance	Anisotropic layer composition and subtraction.

Measured representations, fitting, and sampling

Year	Publication	Authors	Paper	Main idea
1998	Eurographics Workshop on Rendering	Rusinkiewicz	A New Change of Variables for Efficient BRDF Representation	Half-angle / difference-angle coordinates used by many measured and neural BRDFs.
2003	TOG / SIGGRAPH	Matusik et al.	A Data-Driven Reflectance Model	Dense measured BRDFs and low-dimensional material spaces; basis of MERL.
2004	TOG / SIGGRAPH	Lawrence et al.	Efficient BRDF Importance Sampling Using a Factored Representation	Compact factored representation designed for sampling.
2005	EGSR	Ngan et al.	Experimental Analysis of BRDF Models	Comparison of analytic fits to measured isotropic and anisotropic materials.
2015	CGF / EGSR	Dupuy et al.	Extracting Microfacet-based BRDF Parameters from Arbitrary Materials with Power Iterations	Robust extraction of a microfacet distribution from reflectance.
2015	TOG / SIGGRAPH Asia	Nielsen et al.	On Optimal, Minimal BRDF Sampling for Reflectance Acquisition	Sparse acquisition and reconstruction with a learned linear basis.
2018	TOG / SIGGRAPH Asia	Dupuy, Jakob	An Adaptive Parameterization for Efficient Material Acquisition and Rendering	Adaptive acquisition, compact representation, and sampling; associated RGL data.
2018	JCGT	Heitz	Sampling the GGX Distribution of Visible Normals	Exact GGX visible-normal sampling; includes code and errata.
2023	HPG	Dupuy, Benyoub	Sampling Visible GGX Normals with Spherical Caps	Spherical-cap construction for visible-normal sampling.
2025	CGF	Kavoosighafi et al.	Adaptive Sampling for BRDF Acquisition	Adapts measurement count and directions to material complexity.

Neural BRDF representations

These works represent reflectance itself, compress measured data, or provide material priors and sampling. Papers estimating analytic parameter maps from photographs are listed separately below.

Year	Publication	Model / authors	Paper	Representation / operation
2020	CGF / Eurographics	Hu et al. / DeepBRDF	DeepBRDF: A Deep Representation for Manipulating Measured BRDF	Autoencoder latent space for measured BRDF compression, editing, and recovery.
2021	CGF	Sztrajman et al. / NBRDF	Neural BRDF Representation and Importance Sampling	Compact per-material MLP, adaptive angular training, and analytic sampling proxy.
2021	arXiv preprint	Fan et al.	Neural BRDFs: Representation and Operations	Latent-space interpolation and layering with evaluation and sampling.
2023	CGF / Eurographics	Liu et al.	Learning to Learn and Sample BRDFs	Joint meta-learning of BRDF fitting and physical acquisition samples; 2022 preprint.
2024	TOG; presented at SIGGRAPH 2024	Zeltner et al.	Real-Time Neural Appearance Models	Hierarchical neural textures, reflectance decoding, and sampling; broader appearance model, 2023 preprint.
2024	ECCV	Gokbudak et al.	Hypernetworks for Generalizable BRDF Representation	Set encoder and hypernetwork for sparse, unstructured observations of unseen materials.
2025	WACV	Hofherr et al.	On Neural BRDFs: A Thorough Comparison of State-of-the-Art Approaches	Compares analytic-parameter and direct neural representations, including physical properties.
2025	SIGGRAPH Conference Papers	Wu et al.	Neural BRDF Importance Sampling by Reparameterization	Learned integral substitutions for neural BRDF sampling; code.
2026	CGF	Zhou et al. / PBNBRDF	Physically Based Neural BRDF: A Framework for Physically Correct Material Reconstruction, Generation and Editing	Reciprocity by reparameterization, energy control through analytical integration, and chromaticity supervision.
2026	CGF / EGSR	De Oliveira et al.	A Hybrid Neural-Microfacet BRDF Model for Real-Time Rendering	GGX-type component plus a compact neural correction; paper.
2026	Cambridge technical report	Hu	Neural representation of Bidirectional Reflectance Distribution Function	Evaluation, sparse reconstruction, and material synthesis; report based on a 2025 undergraduate dissertation.

Fresnel Term

Fresnel reflectance is a factor in many BRDFs, not a complete BRDF. Schlick's 1994 approximation is:

$$F(c)=F_0+(1-F_0)(1-c)^5, \qquad c=\omega_i\cdot\mathbf h=\omega_o\cdot\mathbf h.$$

For a smooth interface, use the incident cosine relative to its normal; for microfacet reflection, use the half-vector cosine above rather than $\mathbf n\cdot\omega_i$. The exponent 5 is essential. For a non-absorbing dielectric interface with refractive indices $\eta_i$ and $\eta_t$:

$$F_0=\left(\frac{\eta_t-\eta_i}{\eta_t+\eta_i}\right)^2.$$

Schlick is an approximation. Use full dielectric/conductor Fresnel equations for optical accuracy, including total internal reflection, and a thin-film model for interference. See PBRT: Specular Reflection and Transmission and the Belcour–Barla paper above.

BRDF formulas

Equations use native GitHub Markdown math, outside HTML tables. No external math-image service is required. These are common reference forms with explicit conventions, rather than identical transcriptions of every original paper.

Lambert

$$f_r(\omega_i,\omega_o)=\frac{\rho_d}{\pi},\qquad 0\leq\rho_d\leq 1.$$

Phong and Blinn–Phong

Common empirical lobe forms, with $\mathbf r=2\mu_i\mathbf n-\omega_i$ and exponent $s\geq0$:

$$f_{\mathrm{Phong}}=\frac{\rho_d}{\pi}+k_s\max(0,\mathbf r\cdot\omega_o)^s, \qquad f_{\mathrm{Blinn}}=\frac{\rho_d}{\pi}+k_s\max(0,\mathbf n\cdot\mathbf h)^s.$$

These expressions describe the lobe shapes; $k_s$ is an amplitude, not automatically a specular albedo. They are not energy-normalized as written. Original illumination equations and their later normalized BRDF variants should not be mixed.

Cook–Torrance / modern microfacet reflection

With a projected-area-normalized normal distribution function (NDF), the common single-scattering specular form is:

$$f_s(\omega_i,\omega_o)= \frac{D(\mathbf h)G(\omega_i,\omega_o)F(\omega_i\cdot\mathbf h)}{4\mu_i\mu_o}, \qquad \int_{\Omega^+}D(\mathbf m)(\mathbf n\cdot\mathbf m)\,\mathrm d\mathbf m=1.$$

$G$ accounts for masking and shadowing. The denominator and NDF normalization must be used together; see PBRT: Roughness Using Microfacet Theory. Adding $\rho_d/\pi$ independently does not guarantee conservation; multiple scattering and diffuse/specular coupling require care.

For isotropic Beckmann and GGX NDFs, $\alpha>0$ and $\cos\theta_h=\mathbf n\cdot\mathbf h$:

$$D_{\mathrm{Beckmann}}(\mathbf h)= \frac{\exp(-\tan^2\theta_h/\alpha^2)}{\pi\alpha^2\cos^4\theta_h}, \qquad D_{\mathrm{GGX}}(\mathbf h)= \frac{\alpha^2}{\pi\left[\cos^2\theta_h(\alpha^2-1)+1\right]^2}.$$

One common height-correlated Smith choice for isotropic GGX is:

$$\Lambda(\omega)=\frac{\sqrt{1+\alpha^2\tan^2\theta_\omega}-1}{2}, \qquad G_2(\omega_i,\omega_o)=\frac{1}{1+\Lambda(\omega_i)+\Lambda(\omega_o)}.$$

This differs from the separable Smith product $G_1(\omega_i)G_1(\omega_o)$. GGX is an NDF, not a complete BRDF. The ideal smooth limit needs a delta distribution rather than substituting $\alpha=0$ into these expressions.

Ward

For tangent axes $\mathbf t,\mathbf b$, define $h_x=\mathbf h\cdot\mathbf t$, $h_y=\mathbf h\cdot\mathbf b$, $h_z=\mathbf h\cdot\mathbf n$. The original anisotropic Ward form is:

$$f_r=\frac{\rho_d}{\pi}+ \frac{\rho_s}{4\pi\alpha_x\alpha_y\sqrt{\mu_i\mu_o}} \exp\left[-\frac{(h_x/\alpha_x)^2+(h_y/\alpha_y)^2}{h_z^2}\right].$$

For $\alpha_x=\alpha_y=\alpha$, the exponent is $-\tan^2\theta_h/\alpha^2$. This is the BRDF lobe, not an NDF. See the original paper; modified Ward variants have different normalization and grazing-angle behavior.

Oren–Nayar (qualitative approximation)

With roughness $\sigma$ in radians, $a=\max(\theta_i,\theta_o)$, and $b=\min(\theta_i,\theta_o)$:

$$\begin{aligned} f_r&=\frac{\rho_d}{\pi}\left[A+B\max(0,\cos(\phi_i-\phi_o))\sin a\tan b\right],\\\ A&=1-\frac{\sigma^2}{2(\sigma^2+0.33)}, \qquad B=\frac{0.45\sigma^2}{\sigma^2+0.09}. \end{aligned}$$

This widely used approximation is documented in PBRT: Microfacet Models. It reduces to Lambert at zero roughness; see EON for an energy-preserving rough diffuse alternative.

SVBRDF capture and inverse rendering

These papers estimate spatial material properties from images, often jointly with geometry and lighting. Their scene representation (SDF, radiance field, tensor, or Gaussian) is distinct from the BRDF used by the renderer. Generic NeRF and 3D Gaussian Splatting methods without material/lighting decomposition are outside this list.

Year	Publication	Method / authors	Paper	Input / material representation
2015	TOG / SIGGRAPH	Aittala et al.	Two-Shot SVBRDF Capture for Stationary Materials	Flash/no-flash pair; self-similar planar material, anisotropic SVBRDF maps.
2017	TOG / SIGGRAPH	Li et al.	Modeling Surface Appearance from a Single Photograph using Self-augmented Convolutional Neural Networks	Single planar sample under unknown natural lighting; self-augmented learning.
2018	TOG / SIGGRAPH	Deschaintre et al.	Single-Image SVBRDF Capture with a Rendering-Aware Deep Network	Single flash image; normal, diffuse/specular albedo, and roughness maps.
2018	ECCV	Li et al.	Materials for Masses: SVBRDF Acquisition with a Single Mobile Phone Image	Mobile flash photograph; CNN and rendering supervision.
2019	CGF / EGSR	Deschaintre et al.	Flexible SVBRDF Capture with a Multi-Image Deep Network	Variable number of photographs; aggregates complementary observations.
2020	TOG / SIGGRAPH Asia	Guo et al. / MaterialGAN	MaterialGAN: Reflectance Capture using a Generative SVBRDF Model	GAN material prior optimized against input photographs.
2020	CVPR	Li et al.	Inverse Rendering for Complex Indoor Scenes: Shape, Spatially-Varying Lighting and SVBRDF from a Single Image	Single indoor image; material maps and spatially varying illumination.
2021	ICCV	Boss et al. / NeRD	NeRD: Neural Reflectance Decomposition From Image Collections	Image collections; shape, analytic reflectance parameters, and illumination.
2021	TOG / SIGGRAPH Asia	Zhang et al. / NeRFactor	NeRFactor: Neural Factorization of Shape and Reflectance Under an Unknown Illumination	Multiview images; separates geometry, lighting, and reflectance with a learned BRDF prior.
2021	CVPR	Zhang et al. / PhySG	PhySG: Inverse Rendering with Spherical Gaussians for Physics-based Material Editing and Relighting	Multiview images; SDF geometry and spherical-Gaussian reflectance/lighting approximations.
2022	CVPR	Zhang et al. / InvRender	Modeling Indirect Illumination for Inverse Rendering	Unknown static lighting; models indirect illumination for material estimation.
2022	ECCV	Yao et al. / NeILF	NeILF: Neural Incident Light Field for Physically-based Material Estimation	Incident-light field with spatial material parameters.
2023	CVPR	Jin et al. / TensoIR	TensoIR: Tensorial Inverse Rendering	Tensor fields for geometry, surface reflectance, and lighting.
2023	ICCV	Zhang et al. / NeILF++	NeILF++: Inter-Reflectable Light Fields for Geometry and Material Estimation	Joint geometry/material estimation with inter-reflectable incident/outgoing light fields.
2023	TOG / SIGGRAPH	Liu et al. / NeRO	NeRO: Neural Geometry and BRDF Reconstruction of Reflective Objects from Multiview Images	Reflective objects under unknown illumination; geometry and BRDF reconstruction.
2023	SIGGRAPH Asia Conference Papers	Sartor, Peers / MatFusion	MatFusion: A Generative Diffusion Model for SVBRDF Capture	Diffusion prior for photograph-conditioned SVBRDF estimation.
2024	CGF / Eurographics	Luo et al.	Single-Image SVBRDF Estimation with Learned Gradient Descent	Single flash image; learned prior and iterative material-map optimization.
2024	CVPR	Liang et al. / GS-IR	GS-IR: 3D Gaussian Splatting for Inverse Rendering	Gaussian scene representation with explicit material/lighting decomposition.
2025	CVPR	Chung et al.	Differentiable Inverse Rendering with Interpretable Basis BRDFs	Multiview flash images; sparse mixtures of interpretable basis BRDFs on 2D Gaussians.
2025	CVPR	Wu et al. / PBR-NeRF	PBR-NeRF: Inverse Rendering with Physics-Based Neural Fields	Energy-conservation and NDF-weighted specular loss priors for material estimation.
2026	CGF / Eurographics	Iser et al.	High-Gloss SVBRDF Capture Using Bounce Light	Few indirectly lit photographs for high-gloss material capture.

Datasets and implementations

Resource	Link	Use / scope
MERL BRDF database	MERL paper and data entry	100 measured isotropic BRDFs; a standard representation/fitting benchmark.
RGL material database	RGL materials	Measured materials associated with adaptive acquisition, including anisotropic reflectance.
Nielsen et al. data	Code and data	Sparse sampling directions, PCA components, and reconstructed BRDFs.
Two-shot SVBRDF	Code and data	Planar material capture; check the authors' data/code licenses.
NBRDF	Code and pretrained models	Training and rendering measured BRDF networks.
Neural reparameterization sampling	Code	SIGGRAPH 2025 neural BRDF sampler.
NeRFactor	Code	Inverse rendering with a learned reflectance prior.
NeILF++	Code	Geometry/material estimation with inter-reflections.
PBRT	Book, implementation	Reference BRDF/BSDF evaluation, Fresnel, and sampling conventions.

Surveys and further reading

Year	Publication	Paper	Scope
2016	CGF / Eurographics STAR	BRDF Representation and Acquisition	Analytic, measured, and acquisition foundations.
2024	CGF	Deep SVBRDF Acquisition and Modelling: A Survey	Learning-based capture, assumptions, and comparison of SVBRDF methods.
2026	CGF, Early View	Recent Trends in Inverse Rendering	Survey of 2020–2025 inverse rendering; published August 2026.

For an entry point, read Rusinkiewicz and Matusik for measured representations, Walter and Heitz for microfacet conventions, NBRDF and the WACV 2025 comparison for neural representations, and the 2024 SVBRDF survey for image-based capture.

Contributing

Add the publication year, venue/status, an author or publisher link, and a short description of the representation. For image-based methods, state the capture assumptions and whether the method predicts analytic parameters or direct BRDF values. Prefer published versions over duplicate preprints, and label reports/course notes explicitly. Physical-property claims should name the variant and conditions and link to supporting analysis; use native Markdown math outside HTML tables for equations.