✬ ✩

CPSC 3710 Computer Graphics University of Lethbridge

Lighting

• So far, we assumed that each surface has a solid colour (or a blend). The
colour is completely dependent on the surface.
• In reality, colour depends on both the properties of the surface and the
light source.
• A flat surface does not look uniform throughout generally.
• We now look at how to model lighting effects.
• Realistic modelling of lighting requires knowledge in physics: many
approximations and simplifications in computer graphics

✫ ✪
Lighting 1 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Global Lighting

• In “real world”, there can be many light sources.

• Some objects can emit light.
• Other objects can absorb, scatter, and/or reflect light.
• What we see is how much light is scattered and reflected to the viewer
from a particular point.
• When light bounces off a surface, it can then interact with something
else.
• Need to trace light recursively.
• Computationally intensive.

✫ ✪
Lighting 2 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Local Lighting

• To simplify calculations, we compute local lighting.

• For each position and for each light source, compute amount of light
arriving at the viewer.
• Consider only light directly from light sources.

✫ ✪
Lighting 3 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Surfaces

There are three main types of surfaces (can be a mixture of these):

• Specular: most of the light is reflected or scattered in a narrow range of
angles. Appears shiny.
• Diffuse: scattered in all directions. Appears matte or flat.
• Translucent: some light can pass through the surface. Can also bend
light (refraction).

✫ ✪
Lighting 4 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Light Sources

• Light is not simply an intensity. It is an intensity function depending on

wavelength (colour).
• We model a light source with a three component intensity

I = (Ir , Ig , Ib )

representing the intensity of the RGB components of the light.

• There may also be a direction.
• Four basic types: ambient, point, spotlights, distant light.

✫ ✪
Lighting 5 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Ambient Light

• There is often “general” lighting (e.g. sun, overhead lights, etc.) that
provide somewhat uniform lighting in the whole scene.
• Modelling each such source can be computationally intensive.
• Instead, we define a general uniform lighting as the ambient light and
apply this uniformly everywhere:

Ia = (Iar , Iag , Iab )

• Each point in the scene receives the same ambient light.

✫ ✪
Lighting 6 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Point Light Sources

• An ideal point source emits light equally in all direction.

• It is located at a point p0 :

I(p0 ) = (Ir (p0 ), Ig (p0 ), Ib (p0 ))

• For any given point p, the amount of light received from p0 depends on
the distance (squared):
1
i(p, p0 ) = 2
I(p0 )
kp − p0 k

• It is simple but may not be realistic.

• Often bright or dark, but not smooth. Larger light sources can smooth
out shadows, for example.

✫ ✪
Lighting 7 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Spotlights

• Light is projected in a cone. The width of the cone is determined by an

angle.
• Usually light is more concentrated at the center of the cone, and
decreases moving away from center.
• If ~u is unit vector for the direction of the spotlight, and ~v is the unit
vector from spotlight to object, then θ = cos−1 (~u · ~v ) can be used to
determine intensity received.

✫ ✪
Lighting 8 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Distant Light Sources

• We often need to compute a vector from light source to a point.

• If a light source is very far away, this vector can be considered constant
for all points.
• Direction is simply represented as a vector (same for all points).

✫ ✪
Lighting 9 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Phong Lighting Model

• Four vectors are needed to compute the colour of each point p:

– ~n: normal vector to the surface at point p
– ~v : vector from point p to the viewer (center of projection for
perspective viewing)
– ~l: vector from point p to the light source (assuming point light source)
– ~r: direction of a perfectly reflected light from ~l (computed from ~n and
~l)

• All vectors are normalized to have length 1.

• Each light source has three components: ambient, diffuse, and specular,
each with RGB components.

✫ ✪
Lighting 10 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

• For each light source, this can be specified as a matrix:

 
Lra Lga Lba
 
L = Lrd Lgd Lbd 


Lrs Lgs Lbs

• We can use matrices or three different vectors (ambient, diffuse,

specular) in implementation.
• We can also use 4-component RGBA colours.

✫ ✪
Lighting 11 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Phong Lighting Model

• Each point also has a reflective value (in [0, 1]) for each type of light and
colour to determine the proportion of light that is reflected:
 
Rra Rga Rba
 
R = Rrd Rgd Rbd 


Rrs Rgs Rbs

• We can also use 4-component RGBA colours.

• The light seen for each component is simply the corresponding light
source multiplied by the reflection value.
• The amount of light seen is the combination of the three components.
For example, for the red component:
Ir = Rra · Lra + Rrd · Lrd + Rrs · Lrs + Iar

✫ ✪
Iar is an optional global ambient term
Lighting 12 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

• If there are multiple light sources, need to add the contribution from
each light source.
• To simplify presentation, we will remove the r, g, b subscripts

✫ ✪
Lighting 13 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Ambient Reflection

• Each point in the surface has the same reflection coefficient ka ∈ [0, 1].
• I a = k a · La .
• La can be any individual light sources, or global ambient term.
• The RGB components of ka can describe the colour of the surface
(assuming white light)

✫ ✪
Lighting 14 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Diffuse Reflection

• Ideally, diffuse reflector scatters light equally in all directions.

• Also called Lambertian surfaces.
• How much light is scattered depends on the angle between the surface
and light source: brightest if it is perpendicular, darkest if it is parallel:

Rd = kd · cos θ = kd (~l · ~n)

where θ is the angle between ~l and ~n, and kd is some constant (property
of surface)
• The intensity from diffuse term is:

Id = kd (~l · ~n) · Ld

• If we wish to incorporate the distance to the light: divide by a + bd + cd2

✫ ✪
for some constants a, b, c and distance d.

Lighting 15 – 25 Howard Cheng

✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

• If the light source is below the surface, this can be negative (set to 0
instead).

✫ ✪
Lighting 16 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Specular Reflection

• Allows for “shininess”

• Amount of light reflected depends on how close we are to the reflected
angle.
• Drops off quickly as we move further.
Is = ks · Ls cosα φ

• ks is property of the material

• α controls how quickly it drops off (shininess). A value of 100–500
correspond to most metallic surfaces.
• φ: angle between ~r and ~v , so
cos φ = ~r · ~v

• Note: if the dot product is negative, Is = 0 because the light is on the

✫ other side of the surface ✪
Lighting 17 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Blinn-Phong Model

•
Is = ks · Ls (~r · ~v )α
• φ: angle between ~r and ~v .
• ~r has to be computed for each point
• One approximation: use the “halfway vector”:

~ ~v + ~l
H=
k~v + ~lk
~ and increase α (heuristically 4×):
• Replace ~r · ~v by ~n · H
~ 4α
Is = ks · Ls (~n · H)
~ < 0, set Is = 0.
• If ~n · H

✫ ✪
• Why can this be faster? If we assume camera and light are “far away”,

Lighting 18 – 25 Howard Cheng

✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

then ~v and ~l can be assumed to be constant on entire surface.

✫ ✪
Lighting 19 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Normal Vectors Computations

• Sometimes we can just specify the normals by hand.

• If we have a triangle specified by p1 , p2 , p3 in counterclockwise order,
then
~n′ = (p2 − p1 ) × (p3 − p1 )
and
~n′
~n = ′
k~n k
• How does ~n gets transformed by model-view matrix? If M is the
model-view matrix (only the top-left 3 × 3 portion), then transform ~n by
M · ~n does not work (e.g. scaling)!
• Need to transform ~n by the Normal matrix:
N = (M −1 )T
(see textbook for the math behind this)
✫ ✪
Lighting 20 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Gourad vs Phong Shading

• Gourand Shading: determine the colours at each vertex. The fragments

inside the primitive are interpolated from the vertices.
• Can be done in vertex shader.
• Phong Shading: determine the normal vectors at each vertex. The
fragments interpolate the normal vectors and compute colours.
• Can be done in fragment shader.
• If a surface is flat and all vertices have same normal vector, these two
approaches are similar (not exactly the same)
• Otherwise, Phong shading produces more realistic results.

✫ ✪
Lighting 21 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

OpenGL Implementation

• Represent all properties of light and surfaces (Ia , Id , Is , ka , kd , ks as a

vector of RGB components (can also add alpha).
• Specify light position as a 4-diemsnional vector.
• Many libraries will allow componentwise vector multiplication:

Ia ∗ ka = (Iar kar , Iag kag , Iab kab )

• In our matrix/vector library:

– normalize scales a vector to unit length
– dot computes dot product
– cross computes cross product
– Normal computes the normal matrix

✫ ✪
• Pass them as uniform variables to vertex shader.
Lighting 22 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

OpenGL Implementation

In vertex shader:
position = view * model * vPosition;
gl_position = projection * position;
N = Normal * aNormal;
// aNormal passed in as normal in model space
Make the normal vector and position part of the output variable to the
fragment shader (interpolated).

✫ ✪
Lighting 23 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

OpenGL Implementation

In fragment shader:
L = normalize(lightPosition - position)
V = normalize(-position);
// viewer is at origin
H = normalize(V + L);
Light and surface properties can be passed to fragment shader as uniform
variables from application.
Compute ambient, diffuse and specular terms separately and add them.

✫ ✪
Lighting 24 – 25 Howard Cheng
✬ ✩
CPSC 3710 Computer Graphics University of Lethbridge

Ray Tracing

• Ray tracing is a technique to perform global shading computation.

• For each pixel, shoot a ray towards the centre of projection and see what
is the first object it hits. Then bounce it towards a light source (stength
adjusted by angles)
• If the object is reflective, recursively trace the reflected ray (up to some
limit)
• More realistic scenes, can handle shadows and reflections.
• Computationally more intensive.

✫ ✪
Lighting 25 – 25 Howard Cheng