Position-Normal Manifold for Efficient Glint Rendering on High-Resolution Normal Maps

Fujun Luan; Liwen Wu; Milo\v{s} Ha\v{s}an; Ravi Ramamoorthi

REVIEW 4 major objections 6 minor 10 references

Position-Normal Manifold for Efficient Glint Rendering on High-Resolution Normal Maps

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Exact glint NDF evaluation via 4D manifold intersections

desk verdict Genuinely new closed-form P-NDF evaluation and analytical shadow-masking, but the 'exact' label overstates what the renderer actually uses; worth a serious referee with a request for sensitivity analysis. read the letter →

arxiv 2505.08985 v1 pith:3XW7QH6T submitted 2025-05-13 cs.GR

Liwen Wu , Fujun Luan , Miloš Hašan , Ravi Ramamoorthi This is my paper · ORCID

classification cs.GR

keywords glintrenderingnormaldistributionfunctionposition-normalmanifoldmeshintersectionclusterhierarchyshadow-maskingdiffusereflectionsmapfiltering

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that the glint normal distribution function (NDF) of a high-resolution normal map can be evaluated exactly, rather than approximated, by treating the normal-map graph as a 2D manifold in 4D position-normal space. On a piecewise-linear normal map, the P-NDF at a queried normal becomes a sum over triangle intersections, each contributing the footprint kernel at the intersection's texture coordinate divided by twice the triangle's area in normal space. This replaces expensive numerical convolution with simple point-triangle intersections, and makes a cluster hierarchy straightforward to apply, giving roughly an order of magnitude speedup. It also derives an analytical projected-area integral that supplies shadow-masking for specular glints and anti-aliased diffuse reflections from normal maps.

What carries the argument

The load-bearing object is the position-normal manifold: the graph $(u, n(u))$ of the normal map, triangulated into a 4D mesh with barycentric normal interpolation. Evaluating the P-NDF is then projection of this manifold onto the normal plane, implemented as point-in-triangle tests between the query normal $m$ and normal triangles $n(\triangle abc)$, followed by accumulating kernel-weighted inverse-Jacobian contributions. A min-max bounding hierarchy prunes triangles that cannot intersect the query, and a cluster hierarchy approximates large footprints with coarser grids whose normals are fit by weighted least squares. For shadow-masking, Stokes' theorem converts the projected-area integral over each clipped normal triangle into closed-form line integrals over straight edges and ellipse arcs.

What would settle it

Render a specular surface with several normal maps containing large flat regions and isolated tiny bumps, and compare images at epsilon = $10^{-8}$, $10^{-6}$, and $10^{-4}$ at the same sample count; if maximum highlight intensity or glint density changes by more than a few percent, the clamped-pdf claim collapses. A direct check is to compute the true Gaussian-convolved NDF by brute-force binning for such maps and measure the error of the clamped-pdf P-NDF against it.

Watch

Extended reading notes

Core claim

At its center is the claim that Eq. (5) of the paper is an exact solution of Eq. (2) for a piecewise-linear normal map: $$D(m,x) = \sum_{\triangle abc \in k_r} \frac{k_r(a\lambda_0+b\lambda_1+c\lambda_2 - x)\,\mathbf{1}_{n(\triangle abc)}(m)}{2\|n(\triangle abc)\|}.$$ Each normal triangle $n(\triangle abc)$ that contains the queried normal $m$ contributes, and the kernel weight is evaluated at the barycentric texture coordinate of the intersection, normalized by twice the triangle's area in normal space. Because this is a direct evaluation of the un-convolved footprint NDF, the Gaussian micro-roughness convolution required by earlier continuous glint models is avoided; degenerate zero-area triangles are instead handled by clamping the Jacobian to $\epsilon = 10^{-6}$. The same manifold view extends to shadowing-masking: with a piecewise-constant footprint kernel, the projected-area integral over each normal triangle becomes the area of the triangle clipped against a semi-circle and semi-ellipse, and Stokes' theorem converts that area into closed-form line integrals. In short, the paper establishes an exact, mesh-intersection formulation of glint NDFs together with an analytical shadow-masking term.

Load-bearing premise

The load-bearing premise is that replacing triangles whose Jacobian is below epsilon = $10^{-6}$ with a clamped equilateral triangle of exactly that area reproduces the appearance of Gaussian-convolved glints; the paper does not test how sensitive the rendered highlights are to the clamp size or shape.

Editorial extensions

If this is right

At footprint scales near 256 by 256 texels, the full hierarchy brings render time from tens of minutes to about a minute on the test scenes, so the same time budget can buy many more samples per pixel.
Because evaluation no longer requires Gaussian kernel convolutions, arbitrary footprint kernels such as disk or box filters can be substituted; on the isotropic test map, a disk or box filter at half footprint reaches similar NDFs at roughly twice the speed.
The analytical projected-area integral supplies a shadow-masking term for the continuous glint BRDF that prior continuous formulations lacked, improving grazing-angle behavior.
For diffuse surfaces, aggregating the projected area inside the pixel footprint removes aliasing from normal-mapped diffuse reflections at 1 sample per pixel while preserving microstructure detail.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable extension is to treat the clamp epsilon as a user-facing glint sharpness parameter: the same mesh machinery could quantify how glint peak height and sparkle density scale with epsilon, a sensitivity study the paper does not run.
The point-in-triangle formulation maps directly onto ray-tracing hardware; building the normal-triangle mesh as a bounding-volume hierarchy and querying it with a ray along the desired normal could plausibly push glint evaluation to interactive rates.
The analytical projected area could be reused as a building block for multiple-scattering or interreflection corrections on the microsurface, which the paper identifies as an open direction.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The real contribution here is Eq. (5): a closed-form, mesh-intersection evaluation of the un-convolved P-NDF for piecewise-linear normal maps, plus an analytical projected-area derivation for shadow-masking. That is new relative to Yan et al. 2014/2016, which rely on numerical integration or large Gaussian mixtures, and the validation against binning and Monte Carlo (Fig. 8) is solid. The cluster hierarchy is a sensible extension, and the performance numbers, though measured only against self-provided baselines, are consistent with the algorithmic improvement.

The soft spot is the clamping in Sec. 4.1. Eq. (5) is exact only for the singular P-NDF. The actual renderer evaluates Eq. (6), a clamped pdf where triangles with Jacobian below epsilon are replaced by equilateral triangles of area epsilon/2. The paper asserts that clamping is 'just as effective' as Gaussian convolution and shows one qualitative comparison. There is no sensitivity analysis over epsilon, no error metric against the convolved reference, and because epsilon is absolute while Jacobians scale with texture resolution and footprint, the same value may behave differently across scenes. This does not sink the paper, but it does mean the 'exact' claim should be qualified and the authors should be asked to justify the choice of epsilon and tau with actual measurements.

There is a smaller inconsistency in Eq. (9): as written it integrates over d m-tilde but D is a density on the projected hemisphere, so it is missing the 1/m_z factor that appears in Eq. (10). Likely a typo, but it needs fixing. The supplemental derivations for the projected-area integral are detailed and, as far as I checked, coherent.

The citation pattern is fair; the baselines are the authors' prior work, which is appropriate given the direct comparison. The code is promised on GitHub, which is good, though a versioned release would help.

Overall: the mathematical core is sound, the speedup is plausible, and the diffuse shadow-masking application is a nice aside. The clamping equivalence is the main unresolved question, and it is answerable with a sensitivity study. This is a paper for the SIGGRAPH/TOG crowd, and a serious referee can push it into shape. My recommendation: engage, send it to review, and require the epsilon/tau analysis plus the Eq. (9) fix before acceptance.

Referee Report

4 major / 6 minor

Summary. The paper presents a manifold-based formulation for glint rendering on high-resolution normal maps. The authors represent the graph of a continuous normal map as a 4D position-normal triangle mesh and show that P-NDF evaluation reduces to point-triangle intersections, giving Eq. (5) as a closed-form sum that is exact for the piecewise-linear representation. They add a min-max hierarchy and a mesh-clustering hierarchy to accelerate large-footprint queries, and they replace near-singular Jacobian determinants by a clamped value epsilon. The same framework is extended to an analytical projected-area/shadow-masking integral, Eq. (12), with a GGX approximation, and this is applied to anti-aliased normal-mapped diffuse rendering. The results report quality similar to Yan et al. [2014, 2016] with large speedups, especially for large footprint sizes.

Significance. If the claims hold, this is a significant contribution to glint rendering: the P-NDF identity in Eq. (5) is an elegant, parameter-free solution of the continuous formulation, and the reported speedups of roughly one order of magnitude over Yan et al. [2016] are practically important. The analytical projected-area derivation for shadow-masking and diffuse anti-aliasing is a novel extension that fills a gap in prior continuous glint models. The paper is also commendable for releasing code, validating Eq. (5) against binning and Monte Carlo references, and supplying detailed appendix derivations. The main reservations are that the exact identity is not what the renderer actually evaluates, and the two approximations used in practice -- Jacobian clamping and cluster-based approximation -- are validated only qualitatively or against the authors' own no-cluster baseline.

major comments (4)

[Sec. 4.1, Eqs. (5)-(6), Fig. 4] The central exactness claim applies to the unclamped singular P-NDF in Eq. (5), but the actual renderer samples and evaluates the clamped pdf of Eq. (6). The sentence 'an alternative strategy by clamping is just as effective' is not supported by quantitative evidence: no sensitivity analysis on epsilon is reported, no error metric against the Gaussian-convolved reference is given, and Fig. 4 is a single qualitative comparison on one normal map. Since clamping directly sets the height of glint peaks exactly where the Jacobian singularities live, a factor of 10 in epsilon changes peak densities by an order of magnitude. Moreover, epsilon is an absolute threshold while normal-map Jacobians carry units of normal area per texel area, so the same epsilon cannot be expected to behave identically across texture resolutions and footprint scales. The authors should provide an epsilon sensitivity study, compare the clamped pdf against the convolved reference quantitatively, and either justify the absolute threshold or propose a relative or size-aware clamping rule.
[Sec. 4.2, Eqs. (7)-(8), Fig. 15, Tab. 3] The cluster hierarchy changes the P-NDF itself: sampling uses n^l(u/2^l) rather than the original normal map, so the evaluated density is no longer the exact Eq. (5) density. The residual threshold tau is a heuristic chosen per scene (10^-3 in general but 10^-4 for the scratch map in Sec. 5.1), and no error bound connecting the least-squares residual to P-NDF error is given. The ablation in Tab. 3 and Fig. 15 compares clustered against the non-clustered version of the same method, not against the exact Eq. (2) or against the convolved reference, so the claim that the glint pattern is 'well-preserved' is not quantitatively established. I ask the authors to report P-NDF error or rendering error with respect to the unclustered exact evaluation for several tau values and several maps, and to state the tau selection rule independently of the scene.
[Sec. 4.3, Eqs. (9)-(10)] Eq. (9) writes P(omega) as an integral over d\tilde m, while Eq. (10) introduces a 1/\tilde m_z factor when passing to the projected-coordinate measure dm. As written, the two equations are inconsistent in their integration measure: if d\tilde m denotes projected area, the notation should be dm; if d\tilde m denotes solid angle, then a Jacobian factor belongs in Eq. (9). Since the analytical shadow-masking contribution rests on this projected-area derivation, the measure convention must be stated explicitly and used consistently from Eq. (9) onward.
[Sec. 5.1, Tab. 2, Sec. 5.3, Tab. 3] The reported speedups are partly parameter-dependent: the clustering threshold tau is set to 10^-3 for isotropic and brush maps but 10^-4 for the scratch map, and the disk/box-filter experiments in Sec. 5.3 use different footprint sizes than the Gaussian experiments. This makes the headline 'order of magnitude faster' difficult to interpret as a property of the method rather than of the chosen operating points. Please report timing and error for a common parameter configuration, and show how the speedup varies with tau over at least a small grid, so readers can assess the robustness of the comparison.

minor comments (6)

[Fig. 4] The caption 'Binning With convolution' is unclear; please specify which panel is the binning reference, which is the convolved reference, and what the shown error values measure.
[Sec. 4.1, Eq. (3)] The vertex labels n0...n3 are used before their correspondence to u0...u3 is stated; consider adding an explicit equation or sentence defining n_i = n(u_i) immediately before Eq. (3).
[Abstract and Sec. 4.1] The word 'exact' in the abstract conflicts with the clamping step in Eq. (6); suggest hedging the abstract to 'exact for the piecewise-linear normal map up to the singularity treatment'.
[Sec. 5.2] The statement that Yan et al. [2014, 2016] 'simply take the Beckmann shadow-masking using fixed roughness' should specify the roughness value used, otherwise the comparison is not reproducible.
[Sec. 5.3, Fig. 16] The disk/box-filter comparison uses a different footprint scale than the Gaussian case; please state the effective kernel widths explicitly in the text or caption so the comparison is interpretable.
[Appendix A, Eq. (15)] The cluster-domain notation with translated coordinates is difficult to follow; a short worked example for l=1 would help readers verify the least-squares setup.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core manifold-intersection derivation is self-contained; the clamp and GGX approximations are labeled approximations, not predictions.

full rationale

The central derivation, Eq. (5), is obtained directly from the defining delta-integral in Eq. (2) by representing the normal map as a piecewise-linear mesh and summing the kernel-weighted Jacobian contributions over each intersecting normal triangle. This is a change of variables over the same quantity being defined, not a separate fitted input, so the claim that Eq. (5) is an exact solution of Eq. (2) is self-contained. The clamping operation in Sec. 4.1, Eq. (6), replaces near-singular triangles with small equilateral triangles; this is an explicit regularization of the clamped PDF, and the paper states that the clamping is 'just as effective' as Yan et al.'s Gaussian convolution. That assertion is a validation gap or correctness risk, but it is not circular: the clamped PDF is defined independently and compared against the convolved reference in Fig. 4, and no hidden parameter is reused as the target result. The cluster hierarchy in Sec. 4.2 fits cluster normals by weighted least squares (Eqs. 7-8) to approximate the original normal map; this is a clearly labeled approximation with an ablation in Fig. 15 and Tab. 3, not a prediction of the exact P-NDF. The shadow-masking derivation in Sec. 4.3 follows from an exact projected-area integral via Stokes' theorem, and the GGX fit in Eq. (13) is explicitly said to be an approximation fitted to the paper's own analytical projected-area function. Self-citations to Yan et al. [2014, 2016] are used as baselines and as prior formulation context; the correctness of Eq. (5), Eq. (12), and the diffuse BRDF aggregation in Eq. (14) does not depend on those papers' theorems or fitted values. No self-referential theorem is invoked to forbid alternatives, and no known result is renamed as a new derivation. Therefore the paper shows no circularity in its central claim.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central derivation (Eq. 5) is self-contained and does not depend on fitted constants, but the method introduces two hand-set constants (epsilon, tau) that control the approximation quality, plus a fitted GGX surrogate for shadow-masking. The axioms are standard microfacet assumptions; the clamping equivalence is an ad hoc assumption.

free parameters (3)

epsilon = 1e-6
Clamping threshold for Jacobian determinants in Sec. 4.1 (Eq. 6); replaces near-zero normal triangle areas to avoid singularities. The value is chosen by hand and directly controls the peak height of glints; no sensitivity analysis is provided.
tau = 1e-3 to 1e-4
Cluster hierarchy error threshold in Sec. 4.2. Values of 1e-3 or 1e-4 are used per scene (scratch uses 1e-4), so the setting is scene-dependent and affects the trade-off between speed and reconstruction accuracy.
GGX roughness alpha and tangent frame Q = fitted via least squares
Parameters of the smooth GGX projected-area approximation in Eq. (13), fitted to the analytical P(omega) at grazing angles. Used only as a fast approximation for specular shadow-masking, not for the central derivation.

assumptions (5)

domain assumption The normal map is piecewise-linear over each texel's two triangles (barycentric interpolation).
Sec. 4.1 Eq. (3); standard P-NDF model shared with Yan et al. 2014.
domain assumption The microfacet BRDF follows the Smith shadowing-masking model with Lambda(omega) = P(omega)/omega_z - 1.
Sec. 4.3; standard microfacet theory.
domain assumption The footprint kernel k_r is treated as piecewise constant per triangle for the analytical projected-area integral.
Sec. 4.3; approximation enabling the closed-form Stokes integral.
domain assumption Diffuse microfacets are Lambertian with the projection factor 1/n_z used for the surface area measure.
Sec. 4.3, Eq. (14); standard for normal-mapped surfaces.
ad hoc to paper Clamping the Jacobian at epsilon yields a P-NDF visually equivalent to the Gaussian-convolved P-NDF.
Sec. 4.1; asserted with only a qualitative comparison in Fig. 4.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Position-Normal Manifold for Efficient Glint Rendering on High-Resolution Normal Maps." pith.science (2026). https://pith.science/paper/3XW7QH6T

@misc{pith2026250508985,
  author       = {Pith},
  title        = {Pith review of: Position-Normal Manifold for Efficient Glint Rendering on High-Resolution Normal Maps},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3XW7QH6T}},
  note         = {Machine review of arXiv:2505.08985}
}

read the original abstract

Detailed microstructures on specular objects often exhibit intriguing glinty patterns under high-frequency lighting, which is challenging to render using a conventional normal-mapped BRDF. In this paper, we present a manifold-based formulation of the glint normal distribution functions (NDF) that precisely captures the surface normal distributions over queried footprints. The manifold-based formulation transfers the integration for the glint NDF construction to a problem of mesh intersections. Compared to previous works that rely on complex numerical approximations, our integral solution is exact and much simpler to compute, which also allows an easy adaptation of a mesh clustering hierarchy to accelerate the NDF evaluation of large footprints. Our performance and quality analysis shows that our NDF formulation achieves similar glinty appearance compared to the baselines but is an order of magnitude faster. Within this framework, we further present a novel derivation of analytical shadow-masking for normal-mapped diffuse surfaces -- a component that is often ignored in previous works.

Figures

Figures reproduced from arXiv: 2505.08985 by the authors.

**Figure 1.** Our manifold-based P-NDF vs the baselines. Yan et al. [2014, 2016] construct the normal distribution function of a footprint query (P-NDF) and convolve it with a tiny amount of Gaussian roughness, which has no closed-form solution and requires slow numerical approximations. Instead, we show the convolution can be avoided. Our representation converts the P-NDF evaluation to simple manifold intersections, which is an … view at source ↗

**Figure 2.** Our position-normal manifold formulation a) converts the P-NDF integration to finding the manifold projections u𝑖 followed by accumulating a finite number of 𝑘r (u𝑖 −x) | detJ(u𝑖 ) | . In contrast, b) Yan et al. [2014]’s convolved formulation requires computing a complex integral to reason about the NDF. The left images show toy examples of 1D normal and 1D position, and c) shows the full 4D case [PITH_FULL_IMAGE:f… view at source ↗

**Figure 4.** Comparison of P-NDF evaluation. Our analytical evaluation matches the reference given by the binning approach [Yan et al. 2014]. It is also close to Yan et al.’s convolved formulation with small intrinsic roughness (10−4 here). This is in closed form as long as n(u𝑖) = m is solvable. For that purpose, we take a mesh-based manifold representation (Sec. 4.1) that allows easy u𝑖 finding accelerated by mesh clustering (… view at source ↗

Figures from the paper (16 more)

**Figure 3.** Figure 3: Normal map texels are placed on a triangle mesh grid (u0 · · · u3), and barycentric interpolation is used to create the continuous n(u). The right image shows the zoom-in of the dotted region. be easily sampled from 𝐷 by first sampling the footprint kernel u ∼𝑘r (u−x) …

**Figure 5.** Figure 5: Acceleration structures used by our method, shown as a 2D toy example. a) a min-max hierarchy records the normal triangles’ bounding box for every 2 𝑙 ×2 𝑙 spatial region (shown as 2 𝑙 here), which helps prune out the never-intersected triangles. b) a cluster hierarchy…

**Figure 6.** Figure 6: Mesh cluster hierarchy successfully uses fewer triangles to represent the normal map (column 1,3). This works for P-NDF evaluations of both small (column 2) and large (column 4) footprint. Project Intersect [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]

**Figure 7.** Figure 7: Projected area integral domain for each triangle is the intersection of the normal triangle (middle) and 𝝎’s visible normals (top left) on the projected hemisphere (bottom left). Its boundary (right) consists of lines (e.g. n 1n 2 ) and ellipse arcs (e.g. n 0n 1 ). di…

**Figure 8.** Figure 8: Our analytical projected-area integral in Eq. (12) matches the ground truth given by Monte Carlo estimation (column 1-3). It is a low-frequency function so can be reasonably approximated with a GGX projected area function in Eq. (13) (column 4). The numbers show the ro…

**Figure 9.** Figure 9: Continuous glint model vs. other formulations. Chermain et al. approximate the P-NDF using averaged statistics as in LEADR mapping [Dupuy et al. 2013], and Atanasov et al. utilize the discrete glint formulation [Jakob et al. 2014]. Both methods are designed for model…

**Figure 10.** Figure 10: Equal time rendering comparison shows our method (right) is able to use more samples to reduce variance given similar time budget. In contrast, it takes more time for the baseline (left) to obtain less noisy images. The intrinsic roughness smooths the NDF response, so…

**Figure 12.** Figure 12: Shadow-masking comparison on a specular surface suggests our GGX approximation gives a very close rendering compared to the analytical shadow-masking yet is faster. Without correct shadow-masking modeling, Yan et al. [2016] produce darker rendering in grazing angles …

**Figure 11.** Figure 11: Normal maps used in the experiment and their rendering comparison. Our P-NDF (column 2) gives similar rendering as its prefiltered formulation (column 1). The insets show the renderings and the NDFs for different footprint scale (texel numbers per unit footprint) [P…

**Figure 13.** Figure 13: Diffuse appearance developed from our analytical projected area preserves the detailed appearance and surface variation (1st row) of the underlying normal map (insets). With larger query footprint size (2nd row), it resembles the Oren-Nayar BRDF [Oren and Nayar 1994],…

**Figure 14.** Figure 14: Qualitative comparison between normal-mapped and our aggregated diffuse BRDF. When the surface normal is in small (micro) scale, the standard normal mapping method fails to consider each normal texel’s contribution within the image pixel, leading to aliasing artifacts…

**Figure 15.** Figure 15: Qualitative ablation of difference clustering threshold. The glint pattern can be well-preserved when the clustering threshold 𝜏 is selected well (1st and 2nd images). For a large 𝜏, the cluster normals no longer match the ground truth, causing distorted highlights (…

**Figure 16.** Figure 16: Our P-NDF with different footprint kernels. Both the disk and the box filter give similar P-NDFs (insets) and renderings compared to the Gaussian filter. However, they have smaller footprint size thus are faster to compute. The numbers on the images show the inference…

**Figure 17.** Figure 17: Geometric derivation of 𝝎𝑥 m˜ 𝑧 ’s integration. Top: the line integral corresponds to the area 𝑆 ′ , which is the projection of circle arc’s underlying area 𝑆 on m𝑦 axis. Bottom: derivation of the circle’s radius𝑟 and the endpoints’ abscissas 𝑝 𝑖 , 𝑝𝑖+1 on the circle…

**Figure 18.** Figure 18: Qualitative comparison with normal-mapped ground truth rendering. The standard normal mapping requires a very large SPP to capture the glint pattern (1st and 2nd images). Our approach is a more accurate approximation of this normal-mapped ground truth than Yan et al..…

**Figure 19.** Figure 19: Qualitative comparison with Yan et al. [2016] on each normal map. We use a conductor BRDF for renderings with an additional coating layer applied to the flake normal map (4th row) [PITH_FULL_IMAGE:figures/full_fig_p014_19.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 10 canonical work pages

[2016]

that use intrinsic roughness, Additional comparison with Yan et al. [2016]. Figure 19 shows qual- itative comparison with Yan et al. under different footprint scales on scenes in Fig. 9 of the paper. We additionally show renderings using the flake normal map with coating, whose timing is provided in Tab

work page 2016
[5]

journal of graphics tools 10, 3 (2005), 65–71

Mipmapping normal maps. journal of graphics tools 10, 3 (2005), 65–71. Bruce Walter, Sebastian Fernandez, Adam Arbree, Kavita Bala, Michael Donikian, and Donald P Greenberg

work page 2005
[10]

Timing of different stages

and our evaluation without the clustering (ours no cluster) demonstrate faster inference speed compared to their performances on other normal maps. Timing of different stages. Table 5 shows the timing of different rendering stages for the scenes in Fig. 14 of the paper. For indirect bounces, Mitsuba [Jakob 2010] sets their ray differentials to zero that c...

work page 2010
[14]

(26) C Experiment details Entry level of the acceleration structures

Direct BRDF Direct emitter Indirect Total Wrench 0.31 0.14 0.19 0.63 Kettle 0.21 0.12 0.15 0.48 Plate & Cutlery 0.72 0.22 0.10 1.04 The cosine term between the line and the ˜m𝑦 axis is d𝑥 , thus, the line integral of 𝝎𝑥 ˜m𝑧 is: ∮ n𝑖+1 n𝑖 𝝎𝑥 ˜m𝑧d ˜m𝑦 = 𝝎𝑥 2 𝑟 2d𝑦 h arcsin𝑝+𝑝 √︁ 1−𝑝2 i d⊤n𝑖+1/𝑟 d⊤n𝑖/𝑟 . (26) C Experiment details Entry level of the accelerat...

work page 1994
[1967]

Haowen Tan, Junqiu Zhu, Yanning Xu, Xiangxu Meng, Lu Wang, and Ling-Qi Yan

Geometrical shadowing of a random rough surface.IEEE transactions on antennas and propagation 15, 5 (1967), 668–671. Haowen Tan, Junqiu Zhu, Yanning Xu, Xiangxu Meng, Lu Wang, and Ling-Qi Yan

work page 1967
[2005]

JOSA A 22, 11 (2005), 2442–2453

Detailed analytical approach to the Gaussian surface bidirectional reflectance distribution function specular component applied to the sea surface. JOSA A 22, 11 (2005), 2442–2453. Ishaan Shah, Luis E Gamboa, Adrien Gruson, and PJ Narayanan

work page 2005
[2007]

Microfacet models for refraction through rough surfaces. In EGSR. Beibei Wang, Miloš Hašan, Nicolas Holzschuch, and Ling-Qi Yan. 2020b. Example-based microstructure rendering with constant storage. Beibei Wang, Miloš Hašan, and Ling-Qi Yan. 2020a. Path cuts: Efficient rendering of pure specular light transport. ACM TOG (2020). Beibei Wang, Lu Wang, and Ni...

work page 2020
[2021]

InSIGGRAPH 2021 Course: Advances in Real-Time Rendering in Games

Nanite a deep dive. InSIGGRAPH 2021 Course: Advances in Real-Time Rendering in Games . Alexandr Kuznetsov

work page 2021

Show all 10 references

[2022]

IEEE Trans- actions on Visualization and Computer Graphics (2022)

Efficient Specular Glints Rendering With Differentiable Regularization. IEEE Trans- actions on Visualization and Computer Graphics (2022). Zhimin Fan, Jie Guo, Yiming Wang, Tianyu Xiao, Hao Zhang, Chenxi Zhou, Zhenyu Chen, Pengpei Hong, Yanwen Guo, and Ling-Qi Yan

2022
[2023]

arXiv preprint arXiv:2305.02678 (2023)

Real-time neural appearance models. arXiv preprint arXiv:2305.02678 (2023). Shuang Zhao, Lifan Wu, Frédo Durand, and Ravi Ramamoorthi

2023 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [2016]

that use intrinsic roughness, Additional comparison with Yan et al. [2016]. Figure 19 shows qual- itative comparison with Yan et al. under different footprint scales on scenes in Fig. 9 of the paper. We additionally show renderings using the flake normal map with coating, whose timing is provided in Tab

work page 2016

[2] [5]

journal of graphics tools 10, 3 (2005), 65–71

Mipmapping normal maps. journal of graphics tools 10, 3 (2005), 65–71. Bruce Walter, Sebastian Fernandez, Adam Arbree, Kavita Bala, Michael Donikian, and Donald P Greenberg

work page 2005

[3] [10]

Timing of different stages

and our evaluation without the clustering (ours no cluster) demonstrate faster inference speed compared to their performances on other normal maps. Timing of different stages. Table 5 shows the timing of different rendering stages for the scenes in Fig. 14 of the paper. For indirect bounces, Mitsuba [Jakob 2010] sets their ray differentials to zero that c...

work page 2010

[4] [14]

(26) C Experiment details Entry level of the acceleration structures

Direct BRDF Direct emitter Indirect Total Wrench 0.31 0.14 0.19 0.63 Kettle 0.21 0.12 0.15 0.48 Plate & Cutlery 0.72 0.22 0.10 1.04 The cosine term between the line and the ˜m𝑦 axis is d𝑥 , thus, the line integral of 𝝎𝑥 ˜m𝑧 is: ∮ n𝑖+1 n𝑖 𝝎𝑥 ˜m𝑧d ˜m𝑦 = 𝝎𝑥 2 𝑟 2d𝑦 h arcsin𝑝+𝑝 √︁ 1−𝑝2 i d⊤n𝑖+1/𝑟 d⊤n𝑖/𝑟 . (26) C Experiment details Entry level of the accelerat...

work page 1994

[5] [1967]

Haowen Tan, Junqiu Zhu, Yanning Xu, Xiangxu Meng, Lu Wang, and Ling-Qi Yan

Geometrical shadowing of a random rough surface.IEEE transactions on antennas and propagation 15, 5 (1967), 668–671. Haowen Tan, Junqiu Zhu, Yanning Xu, Xiangxu Meng, Lu Wang, and Ling-Qi Yan

work page 1967

[6] [2005]

JOSA A 22, 11 (2005), 2442–2453

Detailed analytical approach to the Gaussian surface bidirectional reflectance distribution function specular component applied to the sea surface. JOSA A 22, 11 (2005), 2442–2453. Ishaan Shah, Luis E Gamboa, Adrien Gruson, and PJ Narayanan

work page 2005

[7] [2007]

Microfacet models for refraction through rough surfaces. In EGSR. Beibei Wang, Miloš Hašan, Nicolas Holzschuch, and Ling-Qi Yan. 2020b. Example-based microstructure rendering with constant storage. Beibei Wang, Miloš Hašan, and Ling-Qi Yan. 2020a. Path cuts: Efficient rendering of pure specular light transport. ACM TOG (2020). Beibei Wang, Lu Wang, and Ni...

work page 2020

[8] [2021]

InSIGGRAPH 2021 Course: Advances in Real-Time Rendering in Games

Nanite a deep dive. InSIGGRAPH 2021 Course: Advances in Real-Time Rendering in Games . Alexandr Kuznetsov

work page 2021