brdf_microfacet — SPECULAR reflectance op

Data kinds: normalmapimage2d

Call: import specularity; specularity.brdf_microfacet(normals, light=(0.0, 0.0, 1.0), view=(0.0, 0.0, 1.0), roughness=0.3, f0=0.04) (or opsspecular.get("brdf_microfacet"))

Usage

GGX / Trowbridge-Reitz microfacet specular BRDF. → (H, W), units 1/sr.

`f_s = D * G * F / (4 (n.l) (n.v))` with

• `D` the Trowbridge-Reitz (1975) / GGX normal distribution,

`a^2 / (pi * ((n.h)^2 (a^2 - 1) + 1)^2) for a = roughness^2`, which

integrates to 1 against `(n.h) dw` over the hemisphere — measured by

20000-point midpoint quadrature in `tests/test_specularity.py` at

relative errors 3.2e-07, 6.3e-08 and 4.0e-09 for roughness 0.2, 0.3 and

0.6, all of it quadrature error (the 200000-point rule gives 3.2e-09,

6.3e-10, 4.0e-11, exactly the 100x a midpoint rule predicts);

• `G` the separable Smith (1967) masking-shadowing term with the GGX

lambda;

• `F` Schlick's (1994) Fresnel approximation,

`f0 + (1 - f0) (1 - v.h)^5`. *f0* is the normal-incidence reflectance:

about 0.04 for common dielectrics, 0.5 to 1.0 for metals. When the exact

Fresnel curve matters, use `match3d.fresnel_reflectance` instead — this

is the approximation the microfacet literature specifies, and it is named

rather than hidden.

Exact ground truth it reproduces: at normal incidence with light, view and

normal aligned, every geometric factor is 1 and the value collapses to

`f0 / (4 pi roughness^4)` in closed form — reproduced bit for bit at

roughness 0.3, 0.5 and 1.0 and to 3.3e-16 relative at roughness 0.1, which

took rewriting the GGX denominator to avoid a cancellation (see the comment

at the code; the textbook arrangement was off by 2.2e-13 relative at the

peak). The lobe is reciprocal in light and view to machine precision

(measured 1.7e-16), and its maximum sits at the half-vector.

`roughness` is the perceptual parameter, squared once to reach the GGX

`alpha` — the convention that makes a linear slider feel linear. A

perfectly smooth surface (`roughness = 0`) is a delta function, not a

finite BRDF, so it is refused rather than returned as an infinity.

Raises `ValueError`: geometry problems as in

:func:brdf_blinn_phong; *roughness* outside `(0, 1]`; *f0* outside

`[0, 1]`.

Detailed usage guide

specular_photometric family guide

References (sample data, literature)

• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).

• Operator provenance and references — the sources of the research/methods this op family came from.

• The canonical algorithm (author, year) and its uses are named in the family usage guide above.

Runnable examples (verified samples that actually call this op)

specular_photometricpy -3.11 examples/specular_photometric.py

Ops the type connects to (they accept image2d as input)

polarization_render

Same category (reflectance)

brdf_blinn_phong · dichromatic_render


*Provenance: specularity.py — SPECULAR operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*

© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.