Study the nature of f things to imaging to -An overvie iew of of - - PowerPoint PPT Presentation

Study the nature of f things

to to imaging

• An overvie

iew of

• f physics-based renderin

ing

Wu u Min Minjie ie Art Art Technic ical l Di Director, Ubi bisoft ft Mo Montreal l Stud Studio io

The origin of PBR In the game industry

Status of PBR

Content

1.What t is P PBR 2.The influenc nce of P PBR

Background knowledge

Seeing is believing?

Seeing is believing? Visible light

Visible light range: 400 nm range: 400 nm –700 nm

What is P PBR

Differen ences es between en PBR and t traditi tion

• nal

l renderin ing

Essence

ce: subject ctiv ive and objecti tive

PBR: Based on the physical properties of the things in the objective world Traditional rendering: Based on the viewer's subjective image

PBR Albedo : 0.04 IOR : 1.635 …… 传统的渲染 Diffuse:黑色 闪亮高光 ……

Style: "process-oriented" vs "result-oriented"

PBR: Decompose a complex phenomenon into a series of formulas and parameters associated Traditional rendering: focus on final result

PBR PBR (Domino effect) Traditional renderin ing (Ic (Iceberg Theory ry)

Function: Function: “all all-weather" weather" vs vs "single "single angle" angle"

PBR: it can always adapt to the environment. Traditional rendering: from a certain perspective, unable to take the overall situation into account

PBR PBR Traditional renderin ing

Details ls of PBR PBR

1.

• 1. Physicall

lly Base ased Lig Lightin ing 2. . Physic icall lly Bas ased Shadin ing 3. . Physic icall lly Base ased Sensit itisin ing

Physicall lly Base ased Lig Lightin ing

Three elements of lighting

Three elements of lighting (to discuss PBL from (to discuss PBL from the perspective of artist): the perspective of artist):

• 1. Intensity
• 2. Color
• 3. Type

Lightin ing g intensit ity

Three common physical units 1.Candela 2.Lumen 3.Lux

Steradian, symbol: Steradian, symbol: sr sr

1. . Unit of solid angle 2.

• 2. Any closed sphere’s solid angle is 4π

Candela, symbol: Candela, symbol: cd cd

• 1. Unit of visible light intensity1/683W/sr

2.

• 2. A common candle emits light with a luminous intensity of roughly one candela.

Lumen, symbol: Lumen, symbol: lm lm

• 1. Unit of luminous flux
• 2. 1 lumen (lm)= 1 cd · sr
• 3. The luminous flux of a common candle is about lumens (220v)

Lux, symbol: Lux, symbol: lx lx

• 1. Unit of luminous fluxIlluminance
• 2. 1 lux= 1 lumen/square meter

Attenua uati tion

Inverse-square law: Light intensity is inversely proportional to the square of distance and attenuates (energy conservation)

Scatter erin ing

Light is forced to deviate from a straight trajectory by one or more paths due to localized non-uniformities in the medium through which it passes.

Mie scatteri ring

Condition o: particle radius >= wavelength of the incident light

Mie scatteri ring g is a as f follows ws:

• 1. Most of the incident lights will scatter along the forward direction
• 2. particle radius will change the model of Mie scattering

Particle ticle radius ius is close close to the wavelen elength gth of

• f the

the incident ident light ht

Particle

ticle radius ius increas reases es

Impact on outdoor natural light intensity

weather Transmissivity Sky light Sunny About 0.85 10000 lux Cloudy About 0.55 1000 lux

IES LIGHT

Photometr

hotometric ic profile profile IES ：illuminating engineering society

IES Light = Maximum intensity (candela) X IES Photometric profile

Lightin ing g color

color temperature symbol:K

Rayleig igh h scatteri ring

Condition: Particle radius <= One tenth of the wavelength of incident light

Rayleig igh h scatteri ring

Scattering

cattering intensity: ensity: inversely ersely proportional portional to the fourth rth power er of the wavelength elength

Lightin ing g type

1.directional light 2.Punctual light 3.area light

Punctua ual l light VS Area light

Specular shadow instance

Punctual ctual light ht Area a light ht

brie rief summary ry of

• f PBL

BL

physically measured value

physical unit

mathematical models and methods

physical laws

Lighting

sun light，sky light，cloud， dust。。。。。。

Physicall lly Base ased Sh Shadin ing

Shading

Shading：material

material response response to to lighting lighting Function

Function：BRDF

BRDF

BRDF BRDF 是什么

Bidirectiona idirectional l Reflectance eflectance Distribution istribution Function unction 1.

• 1. Bidirectional
• 2. Reflectance
• 3. Distribution Function

tgoing radiance (to lens or eye)

Bidirectional

idirectional 向

• 1. The direction from sampling point to camera (eyes)
• 2. The direction from sampling point to point light source

tgoing radiance (to lens or eye)

Reflectance

eflectance 射率

Reflectance = Radiance / irradiance

Irradiance (power/area) ：the power of the light received by current point

Radiance (power/(area x solid angle)) ：the power of the light emitted by current point

tgoing radiance (to lens or eye)

Distribution istribution Function unction

tgoing radiance (to lens or eye)

Picture from ：Naty Hoffman,Background: Physics and Math of Shading

Three different Three different types of BRDF types of BRDF

tgoing radiance (to lens or eye)

Empirical model Physically based model Data-driven mode

Basic principles of physics-based model

• 1. Reciprocal
• 2. Conservation of energy
• 3. Constant positive (Positivity)

Generic Shader Generic Shader

Ci = Ka*ambient() Ci = Ka*ambient() + Kd*diffusion() + Kd*diffusion() + Ks*specular() + Ks*specular()

1.

• 1. Ambient

mbient 2.

• 2. Diffusion

iffusion 3.

• 3. Specular

pecular

SURFACE PRO 3 SURFACE PRO 3

Picture from ：Naty Hoffman,Background: Physics and Math of Shading

Diffusion iffusion Th

The pr process of

• f dif

diffusio ion： 1.

• 1. Refracting into a material
• 2. Scattering in the material
• 3. Scattering out from the material

Picture from ：Naty Hoffman,Background: Physics and Math of Shading

Dif Diffusion model

• 1. Lambert ：Base

Based on

• n smo

mooth ma materia ial l sur urface. Lambert Lambert model model characterist characteristics ics: : ：

SURFACE PRO 3 SURFACE PRO 3 Isotropic (camera view) Light intensity distribution is in line with the cosine law (light angle)

  

lambert

F

i

• lambert

E F    cos   

• 2. Oren-Nayar

derived rived from m Lambert bert model el extended tended to the rough gh surface face controlled ntrolled by roughness ghness(0

(0-1.0)

光滑表面 粗燥表面

Theoret etic ical l basis of O Oren-Na Naya yar model

Based on microfacet model theory Composed of many microfacets Every facet can be seen as a Lambertreflection plane

Oren-Nayar formula 当 σ = 0： A =1，B=0，粗燥因子 = 1

σ ：roughness

what is ρ

ρ = Albedo = Albedo

Two kinds of Albedo texture acquisition

SURFACE PRO 3 SURFACE PRO 3 color checker Cross -polarized lighting with spectralon

Specular Specular:

Cook Cook–Tor

• rra

ranc nce reflect ctio ion model：

microfac acet t theory: y:

1 . The surface face is composed posed microfacet rofacets, , every ry facet et only y does s specular cular reflection lection 2 . 2 2 . Based ed on microfacetnormal rofacetnormal M, every ry facet et only y reflects lects the light ht of single gle direction ection

数学公式：

SURFACE PRO 3 SURFACE PRO 3

) 4(l·n)(v·n () () () F G D F

Torrance Cook

  

 Denominator: Denominator: 4n.l n.lv.n v.n： Correction factor for conversion Correction factor for conversion between micro between micro mirror surface and mirror surface and the overall surface the overall surface Molecule: Molecule: D ：Distribution Distribution function function G ：Geometr Geometry attenuation attenuation function function F ：Fresnel Fresnel function function

Influen ence ces from microst stru ructu ture re of t the material al surface ce:

Roughness:

Value range between 0-1

Square root of slope of facet Half Vector: The

he angle le between ween the hvector ctor bisecting ecting the incident ident light ht I and observation ervation direction ection v. Only when h coincides with the facet’s normal M, the microfacetwill be "activated." Picture from ：Naty Hoffman,Background: Physics and Math of Shading

Distrib ibut ution

• n functio

ionD nD：

Normal distribution probability of activated facets

The performance of distribution function with different roughness (GGX)

Geometr try attenuat ation

• n functio

ion n G：

Distribution probability of facet blocking incident light and reflected light

The performance of geometry attenuation function with different roughness (GGX)

Incident light blocked Reflected light blocked Multiple bounce not be considered

Physical Physical meaning meaning

Picture from ：Naty Hoffman,Background: Physics and Math of Shading

Distribution function(D) Geometry attenuation function(G) Blinn-Phong Beckmann GGX GGXAnisotropic Implicit Neumann Cook-Torrance Kelemen Smith Beckmann GGX Schlick- Beckmann Schlick-GGX

Intuiti

tive ve perform rmanc nce e of p probabil ility ty distrib ibut ution

• n model

Integration Integration of distribution

• f distribution function and geometr

function and geometry attenuation attenuation function: function:

D ×

G

Obtain roughness Obtain roughness measured data measured data

Gloss meterunit：GU

roughness meterunit：μm microfacet microfacet

Anisotropic Anisotropic

Anisotropic Isotropic

Real World: The Microstructure of material surface shows a directional arrangement, more common in artifacts Mathematical model: The distribution of microfacet normal is regular Appearance: The performance of specular is different at different directions, spot shape is stretched Real World: The Microstructure of material surface is irregular, more common in natural things Mathematical model: The distribution of microfacet normal is random Appearance: The performance of specular is same at different directions, spot shape is circular

Specular pecular Occlusion cclusion & Cavity Cavity

Specular pecular Occlusion cclusion

1.

1.To To solve ve the "leakage" akage" problem blem 2. 2.Non Non-PBR PBR "leakage" akage" specular cular occlusion lusion (off) f) specular cular occlusion lusion (on)

Specula lar r Occlusio ion realiza zati tion

• n

1. 1.realization

realization with h AO map 2.Camera, Camera, normal mal direction ection associated

• ciated

3.Controlled Controlled by Roughness ghness Picture and code from "Moving Frostbite to PBR"

When the roughness is minimum, the intensity is 50% of AO intensity When the roughness is maximum and the viewing angle coincides with the normal, the intensity is 100% of AO intensity When the viewing angle is at 90 degrees to the normal, the intensity is zero

Cavity Cavity

Cavity’s role

Simulation of recessed hole formed by microstructure Non-PBR Differences between CavityandSpecular Occlusion:

Small range, only affects the recessed hole formed by microfacetstructure. The intensity is between 0-0.5, irrelevant with roughness Big range, relevant with the object shape and structure. The intensity is between 0-1, controlled by roughness Affects both direct specular and indirect specular Only affect indirect specular

Influen ence ces from the materi rial al :

Metallicity Metallicity：

nonconductor = 0 ，conductor= 1：

When n Metallicity allicity = 1 : 1.

• 1. No

No Diffusion fusion，Only Only Specular cular

• 2. albedo

edo = specular cular color

• r

Reflect ctan ance e and Fresnel el：

F0 F

) 1 ( ) 1 (      F

Reflectance: Reflectance: what is F0： the percentage

centage of specular cular from m the incident ident light ht

the viewing

wing direction(V) ection(V) coincides ncides with h the normal mal(N) (N) and incident ident light(L) ht(L)

How to g get F0 F0：

Non conduction：

based on IOR range:0.02-0.06

F0 F 2 2

OR) (1 ) 1 ( I IOR F   

) 1 ( ) 1 (      F

conduction： index of refraction is variation specular color = = index of refraction range:0.65-0.95

Schlick lick function ction

Fresnel el：

the observation

ervation that t things ngs get more e reflective lective at grazing zing angles. les. F0 F

) 1 ( ) 1 (      F

Fresnel Fresnel function function F：

Variable iable：the the angle( le(R) ) Initial itial value ue ：F0 F0 Description scription reflectance lectance

Fresnel el Reflect ctan ance e Table Table： X

X ：Angle Angle R Y： Reflectance Reflectance

F0 F

) 1 ( ) 1 (      F

F0 ：start point Fresnel Fresnel ：trend trend

Picture from ：Naty Hoffman,Background: Physics and Math of Shading

Porosit ity：

The ratio of the pore's volume to the total volume 范围在0到1之间，即为0到100%

100%之间。 F0 F

) 1 ( ) 1 (      F

Porosity Porosity 's role：

to to descript the influrence from the water

1.

• 1. Darken

ken Albedo edo

• 2. Change

nge Glossiness ssiness 3.

• 3. Boost

st reflectence lectence

Physically Ba Based Sen ensit itising ing Two types of sensor

Two types of sensor ：

• 1. CMOS，CCD

CCD 2.

• 2. Eye

Eye

F0 F

) 1 ( ) 1 (      F

Differe renc nce:

1.

• 1. Angle of view

2.

• 2. Dynamic

Dynamic range range 3.

• 3. Resolution

Resolution

F0 F

) 1 ( ) 1 (      F

Angle of view Angle of view

F0 F

) 1 ( ) 1 (      F

CMOS/CCD EYE the focal length of the lens

Multiple factors:

1.a focal length of approximately 22 mm 2.The overlapping region of both eyes is around 130° 3.central angle of view 双 around 40-60° 4.close to a 50 mm focal length lens

Dynamic ic range ：

the ratio between the largest to smallest possible values of a changeable quantity.

Eye Eye 》CMOS/CCD CMOS/CCD

F0 F

RESOLUT UTIO ION & D DETAIL

CCD/CMOS CCD/CMOS：Symmetrical Eye Eye：Prioritize rioritized d based based on

• n interest

interest，and asymmetrical symmetrical

F F

Gradations

CCD/CMOS Eye - interest

interest

Eye - asymmetrical symmetrical

Symmetrical Symmetrical contrast, sharpness, uniqueness, motion......(red part) the central part of eyes’resolution is much higher than at the edges

Physica call lly Based Sensiti ensitising ng

F F

Gradations

• -》CMOS/

S/CCD CD

F

Gradations

Two procedur ures s to s simulat ate the i imaging ng 1.

• 1. Exposur

ure

• 2. HDR TONE MAPPING

NG

Picture From "Moving Frostbite to PBR"

F

Gradations

Exposur ure

快 门门速 光圈 值 Log EV

) (  快 门门速 （光圈 值光 EV

log 

Exposure Exposure value valueunit

nit ：EV EV camera era settings tings

Exposure Exposure value valueunit unit ：EV EV

the photometric tometric quantity ntity of luminous inous exposure

• sure

F

Gradations

The conversi sion n between en luminan ance/ill llum umina nance ce

EV EV --

• -》luminance

luminance EV EV --

• -》illuminance

illuminance

快 门门速 光圈 值 Log EV

) (  快 门门速 （光圈 值光 EV

log 

EV

E 2 * 5 . 2 

3

2





EV

L

F

Gradations

快 门门速 光圈 值 Log EV

) (  快 门门速 （光圈 值光 EV

log 

EV as an indicator of camera settings

F

• 1. convenie

ient，in intuit itiv ive

• 2. Read

eady for

• r the

the po post t pr proc

• cessin

ing(i. i.e. DOF DOF，motio ion bl blur。。。。。。) )

快 门门速 光圈 值 Log EV

) (  快 门门速 （光圈 值光 EV

log 

The benefit

F

Gradations

快 门门速 光圈 值 Log EV

) (  快 门门速 （光圈 值光 EV

log 

EV & post proces essi sing ng

Aperture --> Vignetting

speed--> motion blur

Aperture --> DOF

F

Gradations

快 门门速 光圈 值 Log EV

) (  快 门门速 （光圈 值光 EV

log 

HDR Tone Mapping

• 1. Display image in 8 bit low dynamic range0-1.0
• 2. Choose the discarded part in high dynamic range
• 3. Keep the original gradient ,contrast,detail as much as possible

Gradations

Two Tone Mapping Curves

• 1. Reinhard
• 2. Filmic

Reinhard

Filmic

1

  x x L

Reinhard VS Filmic

Reinhard VS Filmic

F

Gradations

快 门门速 光圈 值 Log EV

) ( 

Different method： Tri-ace's film simulation:

Film simulation Color Enhancement and Rendering in Film and Game Production: Film Simulation for Video Games Yoshiharu Gotanda tri-Ace, Inc.

The influenc nce of P PBR： 1.

• 1. Essenc

nce e of

• f PBR

PBR

• 2. P

Purpos

• se

e of

• f PBR

PBR 3.

• 3. Signif

ific ican ance e of

• f PBR

PBR

Essence ssence of

• f PBR

PBR

St Stan andardization

Purpose Purpose of

• f PBR

PBR

Aut Autom

• mation

Purpose Purpose of

• f PBR

PBR

As Asse sembly y li line ne

Significance ignificance of

• f PBR

PBR

Mass Mass production, production, reduce costs, improve reduce costs, improve efficiency, quality assurance efficiency, quality assurance

References References

Laurence MEYLAN,tone mapping for high dynamic range images,2006 Chris Wynn,An Introduction to BRDF-Based Lighting,nvidia,2006 Bruce Walter, Stephen R. Marschner, Hongsong Li,Kenneth E. Torrance, Microfacet Models for Refraction through Rough Surfaces Eurographics Symposium on Rendering (2007) Naty Hoffman,Yoshiharu Gotanda,Adam Martinez,Ben Snow,Physically-Based Shading Models in Film and Game Production,SIGGRAPH 2010 joshua pines,color enhancement and rendering in film and game production,SIGGRAPH 2010 sebastien lagarde,Adopting a physically based shading model,2011 Dimitar Lazarov. Physically based lighting in call of duty: Black ops. SIGGRAPH 2011 Course: Advances in Real-Time Rendering in 3D Graphics and Games, 2011. Stephen McAuley. Calibrating lighting and materials in far cry 3. SIGGRAPH 2012 Course: Practical Physically Based Shading in Film and Game Production, 2012. Dan Baker and Stephen Hill. Rock-solid shading - image stability without sacricing detail. SIGGRAPH 2012 Course: Practical Physically Based Shading in Film and Game Production, 2012. Brent Burley. Physically-based shading at disney. part of Practical Physically-Based Shading in Film and Game Production, SIGGRAPH, 2012. sebastien lagarde,laurent harduin,The art and rendering of remember Me GDC2013 David Neubelt and Matt Pettineo,Crafting a Next-Gen Material Pipeline for The Order: 1886 GDC2013 Brian Karis,Specular BRDF Reference,EPIC 2013

References References

Marco Alamia ,Physically Based Rendering - Cook–Torrance Lazanyi, Szirmay-Kalos, Fresnel term approximations for Metals Eric Heitz,Understanding the Masking-Shadowing Function in Microfacet-Based BRDFs,SIG2014 Naty Hoffman,Background: Physics and Math of Shading,SIG2014 Danny Chan,Real-World Measurements for Call of Duty: Advanced Warfare,SIG2015