Light Transport

Christoph Garth
Scientific Visualization Lab
Motivation & Context
 In this chapter, we will follow a principled approach to graphics:

• The visual impression (image) that is recorded by a sensor is created by light

that impacts it.
• Light is emitted by light sources and interacts with objects in the scene
(reflection, scattering, etc.).
• Light travels along rays.

How can we describe and model this?

Which quantities and equations are needed?

Goal: Rendering Equation

Computer Graphics – Light Transport– Motivation & Context 4–1

 specular diffuse

Light is emitted Light travels Light is absorbed / Sensor / eye

at light sources along rays scattered at surfaces captures light

How to quantify light? Flux, Irradiance, Radiance

How to quantify surface illumination Irradiance
How to quantify pixel colors Radiance

Computer Graphics – Light Transport– Motivation & Context 4–2

Radiometric Quantities
 Light shows properties of both, waves and particles.

• wave-optical phenomena: diffraction, interference, polarization

• geometric optical properties: direction, position

Computer Graphics typically neglects wave optics:

• Objects and structures are large relative to the wavelength; diffraction,

interference, polarization are ignored.
• The propagation of light can be modeled geometrically, i.e., by a set of rays;
objects in the scene change the direction of and intensity of light along these
rays.

Computer Graphics – Light Transport– Radiometric Quantities 4–3

 Radiation transport / radiative transfer is used to formulate the change of
“radiation energy” through emission, absorption, reflection and scattering.

This can be expressed by a transfer equation, which we will work towards.

• Photons have the energy E = hν.

• h: Planck constant
• ν: light frequency
• If all material properties are known, the radiation intensity can be computed
with the transfer equation.

Computer Graphics – Light Transport– Radiometric Quantities 4–4

 Rather, light is described as a physical (more precise: radiometrical) unit.

Distinguish:

• Radiometry:
The measurement of electromagnetic radiation.
• Photometry:
The measurement of the optical impression perceived by the human eye.

(In principle, photometry corresponds to radiometry, with the difference that the
spectral response of the eye is respected in the former.)

Computer Graphics – Light Transport– Radiometric Quantities 4–5

 Radiant Energy Q [J]

• Describes the energy carried by photons.

• Assuming wavelength is constant, all photons carry the same energy.

Radiant Flux Φ energy per time [W = Js−1 ]

• Describes the power carried by light, i.e. number of photons per time.
• For a light source, this is the overall number of photons emitted per time.

dQ
Φ=
dt
The notation dQ/dt indicates that integration over t required to obtain Q;
in other words, Q can be found from Φ by integrating over time.

Computer Graphics – Light Transport– Radiometric Quantities 4–6

 Flux Density radiant flux per area [Wm−2 ]

Rate at which flux enters, leaves, or passes through an area; strength of light with
respect to a surface area (independent of direction).

dΦ
E=
dA
Variants:

Irradiance E Radiosity / Radiant exitance B

incoming (incident) flux per area outgoing (reflected + emitted) flux per area

Computer Graphics – Light Transport– Radiometric Quantities 4–7

 Illustration: Irradiance E = dΦ
dA dΦ = EdA

flux from all direc-

tions in the hemi-
sphere above dA

dA
unit area

Irradiance measures the overall radiant flux (light flow, photons per unit time)
into a surface element.

Computer Graphics – Light Transport– Radiometric Quantities 4–8

 How can we understand irradiance E(x) at position x?

• A point x has zero area?!

• Use infinitesimals differential description.

Consider a small amount of flux dΦ(x) incident on a small area dA(x) around position x.
For dA(x) → 0, also dΦ(x) → 0, and the quotient converges to the irradiance at x:

dΦ(x)
E(x) =
dA(x)

Obtaining Φ(x) from E(x) requires integration over surface area.

Computer Graphics – Light Transport– Radiometric Quantities 4–9

 Math Recap: Solid Angle

L A

r
r

The angle of a segment of length L on a The solid angle of an area of size A on a

circle of radius r is sphere of radius r is
L A
α= . Ω= .
r r2
Angle of full circle: Solid angle of full sphere:
2πr 4πr2
α= = 2π [rad = radians]. Ω= = 4π [sr = steradians].
r r2
Computer Graphics – Light Transport– Radiometric Quantities 4–10
 radiant flux per solid angle
W dΦ
Radiant Intensity I sr I= dω

• Radiant flux Φ (light flow, photons per unit time) incident on, emerging from, or
passing through a point in a certain direction.

dA
dω
ω dω

unit solid angle light source Φ

outgoing element
of flux dΦ

The radiant intensity describes the radiant flux of a light source in a specific direction or
the flux that reaches a certain point originating from some object.

Computer Graphics – Light Transport– Radiometric Quantities 4–11

 Inverse Square Law for intensity I

Consider a point light source with radiant intensity I in direction ω.

At distance r, the area subtended by a

solid angle is proportional to r2 .
The number of photons emitted in A
A
direction dω and hitting surface area r
A
dA at distance r (creating irradiance E)
is inversely proportional to r2 . 2r

dΦ dΦ I
⇒ E = = 2 = 2. 3r
dA r dω r

Surfaces appear darker further away from lights.

Computer Graphics – Light Transport– Radiometric Quantities 4–12

 Infinitesimal solid angle and surface area

From which directions does a point p recieve light from an area A? Or, how much light
does A transmit to p?

Solid angle of A seen by p: θ

A
r
Ap A cos θ Ap ≈ cos θA
Ω ≈ =
r2 r2 Ω
p

If an infinitesimal small area dA(x) at position x converges to zero, then so does the solid
angle dω, and the relation
dA(x) cos θx
dω =
rx2
is correct in the limit. We can use this to define the final, most important, quantity.

Computer Graphics – Light Transport– Radiometric Quantities 4–13

 Radiance L flux per solid angle per projected area [ mW2 sr ]

Radiant flux per unit solid angle,

per unit projected area incident surface normal
x φ
on, emerging from, passing Flu
dω
through a surface element in a L= φ θ da
ngle
ωA cos θ li
so
certain direction.
d2 Φ d2 Φ
L = =
dωdAp dωdA cos θ source area projected source area
dA da cos θ

L(x, ω) describes the radiance of a projected (infinitesimal) area at x along am

infinitesimal solid angle dω in direction ω.
Here, d2 Φ indicates that two integrations (dω and dA cos θ) are needed to obtain a non-infinitesimal value Φ.

Computer Graphics – Light Transport– Radiometric Quantities 4–14

 Conservation of Radiance – Why is radiance the key quantity for Computer Graphics?

Outgoing flux from dA1 into direction r:

A1
dω1 = dA2 cos θ2 · r−2 A2

d2 Φ1 = L1 · dA1 cos θ1 · dω1

θ1 θ2
= L1 · dA1 cos θ1 · dA2 cos θ2 · r−2 dA1 r
dA2
Incoming flux to dA2 from direction −r: dω2
dω1
2
d Φ2 = L2 · dA2 cos θ2 · dω2
= L2 · dA2 cos θ2 · dA1 cos θ1 · r−2

Flux is conserved (dΦ1 = dΦ2 ) if and only if radiance (L1 = L2 ) is.

Computer Graphics – Light Transport– Radiometric Quantities 4–15

 Recap – Radiometric Quantities

type expression symbol units name

energy Q [J] radiant energy

power/flux dQ
W = Js−1
 
dt Φ radiant flux

flux density dQ
Wm−2
 
dAdt E irradiance

flux density dQ
Wm−2
 
dAdt B radiosity

radiant density dQ
 W 
dAΦ dωdt
L m2 sr
radiance

intensity dQ
W
dωdt I sr radiant intensity

Computer Graphics – Light Transport– Radiometric Quantities 4–16

 In the following:

In order to come up with a transport equation for light, let us assume the
following:

• Light in a scene is emitted by objects (light sources).

• Between objects in a scene there is a vacuum, i. e., light propagates through
empty space without manipulation; radiance is conserved.
This assumption is typical in computer graphics, however does not take account for scattering effects.

For the sake of clarity, these will be covered later.

• Light is reflected at surfaces depending on the material of that surface.

• The image of the scene is produced by computing radiance at a sensor.

Let’s consider (briefly) light sources and surface reflection next.

Computer Graphics – Light Transport– Radiometric Quantities 4–17

Light Sources
 Types of Light Sources

Point lights
• same intensity in all directions
• specified by position

Directional (parallel) lights

• e. g. the sun
• specified by direction

Spot lights
• restricted direction by angle range
• specified by position, direction and opening angle

Computer Graphics – Light Transport– Light Sources 4–18

 Illustration: Effects of different light sources

point light spotlight area light

intensity intensity / angle radiance / geometry

Computer Graphics – Light Transport– Light Sources 4–19

 Area lights
• light sources with (finite) dimension
(think: a continuum of point lights)
• simulation is costly ( Chapter 7 – Path Tracing)

penumbra

umbra

Computer Graphics – Light Transport– Light Sources 4–20

Surface Reflection
 The reflection of light at surface points is described by the
bidirectional reflectance distribution function (BRDF).
Fraction of radiance reaching the surface
from inf. solid angle dωi , reflected into z

direction ωr : ωi
ωr

dLr (ωr ) θi θr
fr (ωi , ωr ) = y
Li (ωi ) cos (θi ) dωi

Often described using four angles: φi

φr
• θi , θr – angles to the normal x

• φi , φr – angles around the normal

Computer Graphics – Light Transport– Surface Reflection 4–21

 BRDF properties:

In general, a BRDF is a 6-dimensional “object”:

• 2× positions on the surface, 2 × 2 angle parameters: fr (x, ωr , ωi )

Often, BRDFs are assumed to be constant (in x) per object, and an isotropic material
presumed.

• An isotropic BRDF is invariant under rotation around the normal vector φi and φr
can be omitted. This yields a simplified description with only 3 parameters.

BRDFs have to be formulated to reproduce the physics of surface reflection.

Computer Graphics – Light Transport– Surface Reflection 4–22

 fr is really a distribution, in the mathematical sense, similar to a probability density. It
can assume “values” from 0 (total absorption) to ∞ (perfect mirror reflection). These
really only make sense if integrated.

Physics informs the following additional conditions:

Helmholtz reciprocity: A ray of light behaves symmetrically under reversion.

fr (x, ωr , ωi ) = fr (x, ωi , ωr )

Conservation of Energy: The reflected radiance at point is bounded by the incident

radiance over all directions (Ω+ ).
Z
fr (x, ωr , ωi ) cos θi dωi ≤ 1 ∀x, ωr
Ω+

Computer Graphics – Light Transport– Surface Reflection 4–23

Rendering Equation
 Combining emission and surface reflection:

Kajiya & Immel (1986): The outgoing radiance at point x in direction ωr is given by:
emission reflection
z }| { z }| {
L (x, ωr ) = Le (x, ωr ) + Lr (x, ωr )
Z
= Le (x, ωr ) + f (x, ω , ω ) L (x, ω ) cos θi dωi
| {z } |r {zr i} | i {z i} | {z }
emission Ω BRDF irradiance orientation

This rendering equation is the foundation of physically-based rendering. Many

rendering approaches concentrate on single parts of this equation.

Conservation of radiance allows us to connect incoming light at a surface point to

outgoing light at all other (visible) surface points, all the way to the sensor.

Computer Graphics – Light Transport– Rendering Equation 4–24

 Unfortunately, solving the rendering equation analytically is usually not possible
in practical applications.

During the course, we will cover different approaches for approximating the
rendering equation.

• Chapter 5: Surfaces
• local illumination models
• BRDFs for different materials
• simplified BRDFs and subsurface scattering (BSSDFs)
• Chapter 6: Ray Tracing
• direct approximation by tracing light rays
• Chapter 7: Global Illumination
• probabilistic approximation using Monte-Carlo integration
• approximation using finite element method (FEM)

Computer Graphics – Light Transport– Rendering Equation 4–25

 BRDF BSSDF
Outlook:
Enhancement with subsurface
scattering

BRDF (only reflection) BSSRDF (subsurface scattering)

Computer Graphics – Light Transport– Rendering Equation 4–26

Global Illumination
 One part of (photo-realistic) rendering is the modeling of how the light in a scene
spreads until received by the sensor (eye or camera).

Some influencing factors are:

• light sources
• lighting models
• material properties
• orientation of surfaces (surface normals)

How do these influence the propagation of light?

Computer Graphics – Light Transport– Global Illumination 4–27

 Local illumination

^
Only respects direct illumination (reflection) by
(point) light sources, independent of other objects
in the scene.
• e. g. no shadows

Global illumination
Additionally, includes indirect effects.

^
• multiple reflections
• transmissions
• shadows

Computer Graphics – Light Transport– Global Illumination 4–28

 Example: local illumination only

ht
direct lig

direct lig
ht

Computer Graphics – Light Transport– Global Illumination 4–29

 Example: local illumination with shadows

ht
direct lig

Computer Graphics – Light Transport– Global Illumination 4–30

 Example: global illumination

ht
direct lig

ce
un
bo
st
fir
seco
nd b
ounc
e

Computer Graphics – Light Transport– Global Illumination 4–31

 Example: global illumination

Computer Graphics – Light Transport– Global Illumination 4–32

 Physically-based Rendering
• Description of light with physically correct units.
• Reflection models simulate the physics of materials.
• Numerical simulations - realizing radiation transport including visibility.
• visibility: light is transported only between two points that can “see” each other

Result: light hitting the sensor from the points visible to it, illuminated directly
and indirectly

Computer Graphics – Light Transport– Global Illumination 4–33

 Refractive and reflective caustics are not reproduced with local illumination.

Computer Graphics – Light Transport– Global Illumination 4–34

 Indirect illumination (“color bleeding”) is not reproduced with local illumination.

Computer Graphics – Light Transport– Global Illumination 4–35

 scattering scattering & diffraction

Scattering in participating media is not reproduced with local illumination.

Computer Graphics – Light Transport– Global Illumination 4–36

Recap & Outlook
 In this chapter, we have considered

• the physical nature of light, and physical quantities to describe it;

• radiance as the primary quantity considered and computed in graphics;
• different types of light sources;
• the bidirectional reflectance distribution function (BRDF) as a generic way to
characterize the reflection of light at surfaces;
• the rendering equation as the fundamental equation for image synthesis.

In the next chapter, we will look at materials more closely ( standard BRDFs),
and then proceed to a fundamental rendering algorithm – ray tracing.

Computer Graphics – Light Transport– Recap & Outlook 4–37