Modeling and rendering of layered materials (다층 재질의 모델링 및 렌더링)

재질감 표현을 위한 렌더링 기술 동향

Modeling and Rendering of Layered Materials

이 주 행ETRI

한국CAD/CAM학회 이론 및 응용 연구회2010년 3월 26일 (금)

Agenda

•Motivation / Background / Previous Works

•Analysis of Previous Method: W2L Model

• Proposed Method / Results

• Summary / Q & A

Motivation

•Layered materials are ubiquitous

- Simple, versatile and effective

- Suitable for digital prototyping application

•A trial to generalize the previous work: observation & experiment

Background

•Digital modeling of material appearance

• Rendering in digital image synthesis

Previous Works

•Kubelka and Munk 1931, Hanrahan and Krueger 1993

•Neumann and Neumann 1989

•Kelemen and Szimay-Kalos 2001

• Schlick 1993, Lafortune et al 1997

•Weidlich and Wilkie 2007, 2009

Evaluation Criteria

•Closed mathematical form vs. simulation

•Handling of physical behaviors: (ex) scattering, absorption, internal reflection, micro-facets, Fresnel

• Integration with rendering algorithms: (ex) unbiased MC rendering requires sampling PDFs and a way of quality control

Overview of W2L

W2L Model

• [Weidlich and Wilkie 2007] = W2L Model

• "Arbitrarily layered micro-facet surfaces", GRAPHITE 2007 + SIGGRAPH ASIA 2009 Course

ExamplesThe final classification can be seen in figure 4.4, although it should be noted that this is probably not

an exhaustive list of what can be achieved by this technique. Note that there are 8 instead of 6 types.

One represents the further split into tinted and non-tinted varnish that can be performed for all types,

and one stands for the group of materials that consists of more than two layers, i.e. multi-layered

materials.

a)Glossy Paint

c)Frosted Paint

d)Metal Foil

e)Metallic Paint

f)Frosted Metal

g)Patina

b)Tinted Glazing

h)Multi-Layer

Torrance-Sparrow

Diffuse

Smooth

Metal

Coloured Solid

Interfaces: Materials:

Colourless Solid

Tinted Varnish

Clear Varnish

Figure 4.4: Examples of various surface types that can be generated by using our layered model in different configurations.

In order to properly distinguish the various cases the icons do not exhibit the simplifying assumption shown in figure 3.1

which we use for all our actual BRDF computations. The micro-facets are much smaller than the layer thickness in this

drawing.

28

[ Weidlich and Wilkie 2007 ]

Comments on W2L Model

•With a few control parameters,

• it can represent a various types of materials

•with a closed form BRDF model, and

• rendering results are physically plausible.

Physical Plausibility

•Application of physically correct micro-facet model: Torrance-Sparrow

•Application of Beer-Lambert Law to compute absorption assuming that coating is a non-scattering medium, which seems not correct.

•Consideration of total internal reflection (TIR). Actually, ineffective.

Closed Form Model

•Geometric simplification of scattering and reflection

•No additional ray-object intersection test guarantees performance = only a single ray path is considered at each intersection

•We argue it as a over-simplification that limits handling more general cases

Well...

• Basically, W2L is a nice approach to handle layered materials in CG

But,

•We need to verify its building blocks as well as basic assumptions.

Experiments with W2L

Micro-faceted Coating

roughness 1 0.1 0.01 0.001 0.0001

3.5

1.5

ior

Direct reflection from coating layer

Absorption inside a Coating

Beer-Lambert Law

•Foundation of spectroscopy

•Beer-Lambert law is commonly used in spectroscopy to derive the absorption coefficient of non-scattering media based on linear attenuation.

• Linearity breaks by the scattering of light due to particulates of the sample or in a dense media

•A modified Beer-Lamber law for a scattering media such as a biological tissue is available [Delpy et al 1988, Sassroli and Fantini 2004]

• But, too costly to evaluate the modified Beer-Lamber law

Total Internal Reflection

•TIR = (1-G) + T*G = (1-G) + (1-F)*G = 1 - F*G

• The range of G is [0.94,1] when IOR < 1.41.

•Otherwise, G=1.

•Hence, the effect of G in TIR is not noticeable.

G

Genelization of W2L

Basic Idea of W2L

View direction wo is given.

V

Get wor from Snell's law

V

Get wir from pdf sampling from coating

V

Get wi from Snell's law

V

Get incoming radiance L from wi

LV

LV

Get BRDF of coating layer, absorption, and attenuation

LV

In W2L, everything happens between a pair of "big" micro-facets!

Generalization

View direction wo is given.

V

Get wor from Snell's law

V

Get wir from pdf sampling from coating

V

Note (1): Micro-facets are much smaller than the coating depth.

V

Note (1): Micro-facets are much smaller than the coating depth.

V

Get wi from Snell's law

V

Get incoming radiance L from wi

LV

LV

Get BRDF of coating layer, absorption, and attenuation

Note (2): A ray path can be further subdivided to approximate scattering

V

V

Note (2): A ray path can be further subdivided to approximate scattering

V

Note (2): A ray path can be further subdivided to approximate scattering

V

Note (2): A ray path can be further subdivided to approximate scattering

LV

For implicit ray bundles, get BRDF, absorption, and attenuation in a MC way

Generalization of W2L = GW2L

•More general configuration of micro-facets to decouple normals

• Implicit ray bundles to approximat scattering in multiple paths

Experiments with GW2L

The most promising trial so far.

Rendering Setups

•PBRT-v2 for rendering

•Metropolis / IGI

• Extension through custom plug-ins in C++

•Coating layer has no diffuse/ambient terms to be more physically acurate

cfg_1 = W2L

cfg_3 = GW2L

cfg_1 cfg_3

base +

coating

coating

base

cfg_1

b+c

cfg_1

b

cfg_3

b+c

cfg_3

b

ior=1.5ior=1.5 ior=3.5ior=3.5

d=1 d=3 d=1 d=3

ior=1.5ior=1.5 ior=3.5ior=3.5

d=1 d=3 d=1 d=3

Effect of Implicit Ray Bundles

b+c

b

nb 1 10 100

time 1:46 2:40 11:50

Does it have an analytic form?

Can it be baked?

Results

glossy paint frosted paint tinted glazing metal foil metallic paint frosted metal patina multi-layer

coatingcolor

coating rough

base material

base color

baserough

tinted tintedtintedtintedtinted

rough roughrough

solidsolidsolid metalmetalmetalmetal

tintedtinted white

diffusediffusediffuse rough rough

glossy paint frosted paint tinted grazing metallic foil metallic paint frosted metal patina

Summary

•Analysis of W2L model

•Generalization of W2L model as GW2L

• Presentation of experimental results

Future Works

•Verification of GW2L algorithm

•Derivation of analytic form for GW2L

•Baking or GPU-based acceleration for re-shading/re-lighting

•Comparison with “real” coated materials

Q & A

Joohaeng@etri.re.kr