Photo-realistic Rendering and Global
Illumination in Computer Graphics
Spring 2012
Visual Appearance
K. H. Ko
Department of Mechatronics
Gwangju Institute of Science and Technology
Shadows

Shadow algorithms determine which surface
can be “seen” from the light source.
 They
are essentially the same as visible-surface
determination algorithms.
 The surfaces that are visible from the light source
are not in shadow.
 Those that are not visible from the light source are
in shadow.

First we consider shadow algorithms for point
light sources.
2
Shadows

When a point on a surface cannot be seen from a
light source, then the illumination calculation must
be adjusted to take it into account.
I   I a k a O d 


S i f att i I p  i k d O d  ( N  L i )  k s O s  ( R i  V )
n

1 i  m
 Si:
0, if light i is blocked at this point. 1, if light i is not
blocked at this point.


Note that areas in the shadow of all point light
sources are still illuminated by the ambient light.
For simplification, we assume that all objects are
polygons.
3
Scan-Line Generation of Shadows



One of the oldest method:
Augment a scan-line algorithm to
interleave shadow and visiblesurface processing.
Using the light source as a center
of projection, the edges of
polygons that might potentially
cast shadows are projected onto
the polygons intersecting the
current scan line.
When the scan crosses one of these
shadow edges, the colors of the
image pixels are modified
accordingly.
4
A Two-Pass Object-Precision shadow
Algorithm


It performs shadow determination before visible-surface
determination.
It processes the object description by using the same algorithm
twice:




For the viewpoint.
For the light source.
The results are then combined to determine the pieces of each
visible part of a polygon that are lit by the light source and the
scene is scan-converted.
The shadows are not dependent on the viewpoint.

All the shadow calculations may be performed just once for a series of
images of the same objects seen from many different viewpoints as
long as the light source and objects are fixed.
5
A Two-Pass Object-Precision shadow
Algorithm

Determine those surfaces that are visible
from the light source’s viewpoint.



The output of this pass is a list of lit polygons.
The lit polygons are transformed back
into the modeling coordinates and are
merged with a copy of the original
database as surface-detail polygons,
creating a viewpoint-independent merged
database.
Hidden-surface removal is then
performed on a copy of this merged
database from the viewpoint of an
arbitrary observer.
6
A Two-Pass Object-Precision shadow
Algorithm
7
A Two-Pass Object-Precision shadow
Algorithm
8
A Two-Pass Object-Precision shadow
Algorithm

Multiple light sources can be handled by processing
the merged database from the viewpoint of each new
light source, merging the results of each pass.
9
Shadow Volumes


A shadow volume is defined by the light source and an object,
and is bounded by a set of invisible shadow polygons.
Shadow polygons are not rendered themselves,
but are used to determine whether the other
objects are in shadow.
10
A Two-Pass z-Buffer Algorithm

A shadow-generation method based on two passes through a
z-buffer algorithm



For the viewer
For the light source
It is based on image-precision calculations.
11
Global Illumination Shadow
Algorithms


Ray-tracing and Radiosity algorithms
They have been used to generate some of the most
impressive pictures of shadows in complex
environments.
12
Transparency

Much as surfaces can have specular and diffuse
reflection, those that transmit light can be transparent
or translucent.
 Transparent materials: We
can see clearly through them
although in general the rays are refracted.
 Diffuse transmission occurs through translucent materials
such as frosted glass.

We consider here two things: non-refractive
transparency and refractive transparency
13
Non-refractive Transparency

Simple model: Ignores refraction.
 Light
rays are not bent as they pass through the surface.
 Note that total photographic realism is not always the
objective in making pictures.

Two methods that approximate the way in which the
colors of two objects are combined when one object
is seen through the other.
 Interpolated transparency
 Filtered transparency
14
Non-refractive Transparency

Interpolated Transparency
x
 Determines
the shade of a pixel in the
2
intersection of two polygons’ projections
by linearly interpolating the individual
1
shades calculated for the two polygons
 Iλ = (1-kt1)Iλ1 + kt1 Iλ2.
z
Line of sight
 kt1: the transmission coefficient. It measures the
transparency of polygon 1. When 1, perfectly transparent.
 1- kt1 : the polygon’s opacity.
 For a more realistic effect, interpolate only the ambient and
diffuse components of polygon 1 with the full shade of
polygon 2, and then add in polygon1’s specular
component.
15
Non-refractive Transparency

Filtered Transparency.
 It
treats a polygon as a transparent filter that selectively
passes different wavelengths.
 Iλ = Iλ1 + kt1 Otλ Iλ2
 Otλ is polygon 1’s transparency color.
16
Refractive Transparency

More complex to model since the geometrical and
optical lines of sight are different.
Indices of refraction
Snell’s Law
17
Refractive Transparency

Calculating the refraction vector

T  
(N  I ) 
r

1   r  (1  ( N  I ) ) N 
2
2
r
I
18
Refractive Transparency

Total Internal Reflection
 When
light passes from one medium into another whose
index of refraction is lower, the angle θt of the transmitted
ray is greater than the angle θi.
 If θi becomes sufficiently large, then θt exceeds 90o and the
ray is reflected from the interface between the media,
rather than being transmitted.
 The smallest angle θi of at which it occurs is called the
critical angle.
 When sinθt is set to 1, θi = sin-1(ηtλ / ηiλ )
19
Global Illumination

An illumination model computes the color at a point
in terms of light directly emitted by light sources and
of light that reaches the point after reflection from
and transmission through its own and other surfaces.
 Global
illumination: indirectly reflected and transmitted
light
 Local illumination: light that comes directly from the light
sources to the point being shaded.
 So far global illumination has been modeled by an ambient
illumination term.

That was held constant for all points on all objects.
20
Global Illumination


Much of the light in real-world environments does
not come from direct light sources.
Two different classes of algorithms for global
illumination
 Ray

tracing: view-dependent
Given the viewer’s direction, discretize the view plane to determine
points at which to evaluate the illumination equation.
 Radiosity: view-independent

Discretize the environment and process it in order to provide
enough information to evaluate the illumination equation at any
point and from any viewing direction.
21
Recursive Ray Tracing

The basic ray-tracing algorithm for visible-surface
determination is extended to handle shadow,
reflection and refraction.
 It
determines the color of a pixel at the closest intersection
of an eye ray with an object, by using any of the
illumination models.
 To calculate shadow, we fire an additional ray from the
point of intersection to each of the light sources.

If one of these shadow rays intersects any object along the way,
then the object is in shadow at that point and the shading algorithm
ignores the contribution of the shadow ray’s light source.
22
Recursive Ray Tracing

Extended illumination model for ray tracing model to include
specular reflection and refractive transparency
I   I a k a O d 


S i f att i I p  i k d O d  ( N  L i )  k s O s  ( R i  V )
n
 k
s
I r  k t I t
1 i  m
 Irλ


is the intensity of the reflected ray, kt is the transmission coefficient,
and Itλ is the intensity of the refracted transmitted ray.
Values for Irλ and Itλ are determined by recursively evaluating the
equation at the closest surface that the reflected and transmitted rays
intersect.
To approximate attenuation with distance, Iλ calculated for each ray is
multiplied by the inverse of the distance traveled by the ray.
23
Recursive Ray Tracing

Extended illumination model for ray tracing model to
include specular reflection and refractive
transparency
I   I a k a O d 
S
i

f att i I p  i k d O d  ( N  L i )  k s O s  ( R i  V )
n
 k
s
I r  k t I t
1 i  m
 Si
is made as a continuous function of the kt of the objects
intersected by the shadow ray is used.

A transparent object obscures less light than an opaque one at those
points it shadows.
24
Recursive Ray Tracing

In addition to shadow rays, recursive ray-tracing
algorithm conditionally spawns reflection rays and
refraction rays from the point of intersection.
 The
shadow, reflection, and refraction rays are often called
secondary rays to distinguish them from the primary rays
from the eye.
25
Recursive Ray Tracing

Each of the reflection and refraction rays, in turn,
recursively spawn shadow, reflection, and refraction
rays.

Ray-tracing is particularly prone to problems caused
by limited numerical precision.
 False
intersection can result in visual problems
26
Recursive Ray Tracing
27
Recursive Ray Tracing

Efficiency Considerations

Item buffers


Reflection maps


Ray is not cast if its contribution to the pixel’s intensity is estimated to be
below some preset threshold.
Light buffers


Do less work for the secondary rays than for primary rays
Adaptive tree-depth control


Not to use it at all when determining those objects directly visible to the
eye.
Shadow rays are special in that each is fired toward one of a relatively
small set of objects. Increase the speed with which shadow rays are
processed. Similar to 3D spatial partitioning of the 3D view of its light.
Ray classification

Spatial-partitioning approach.
28
Recursive Ray Tracing

A Better Illumination Model
 The
specular light expressions are scaled by a wavelengthdependent Fresnel reflectivity term.
 Also take into account the contribution of transmitted light
directly emitted by the light sources.
 Is also scaled by the Fresnel transmissivity term.
 The global reflected and refracted rays take into account
the transmittance of the medium through which they travel.
29
Recursive Ray Tracing

Area-Sampling Variations
 One
of conventional ray tracing’s biggest drawbacks is that
this technique point samples on a regular grid.
 To avoid aliasing resulting from point sampling on a
regular grid by casting solid beams rather than infinitesimal
rays.



Cone tracing: generalizes the linear rays into cones.
Beam tracking: an object-precision algorithm for polygonal
environments that traces pyramidal beams, rather than linear rays.
Pencil tracing: approach to solve some of the problems of cone
tracing and beam tracking.
30
Recursive Ray Tracing

Distributed Ray Tracing
 It
is based on a stochastic approach to supersampling that
trades off the objectionable artifacts of aliasing for the less
offensive artifacts of noise.
 The ability to perform antialiased spatial sampling can also
be exploited to sample a variety of other aspects of the
scene.

Motion blur, specular reflection from rough surface, etc.
 “Distributed” means
that rays are stochastically distributed
to sample the quantities.
31
Recursive Ray Tracing

Stochastic sampling
 Aliasing
results when a signal is sampled with regularly
spaced samples below the Nyquist rate.
 If the samples are not regularly spaced, the sharply defined
frequency spectrum of the aliases is replaced by nose, an
artifact that viewers find much less objectionable than the
individually recognizable frequency components of regular
aliasing such as staircasing.
32
Recursive Ray Tracing

Stochastic sampling
 Pure
random samples may cluster together in some areas
and leave others unsampled.

Use of a minimum distance Poisson distribution in which no pair of
random samples is closer than some minimum distance.
 Satisfactory approximation
to the minimum-distance
Poisson distribution

Displace the position of each element of a regularly spaced sample
grid by a small random distance.
33
Recursive Ray Tracing

Sampling other dimensions





The same basic technique of stochastic sampling can also be used to
distribute the rays to sample other aspects of the environment.
Motion blur is produced by distributing rays in time.
Depth of field is modeled by distributing the rays over the area of the
camera lens.
The blurred specular reflections and translucent refraction of rough
surfaces are simulated by distributing the rays according to the specular
reflection and transmission functions.
Soft shadows are obtained by distributing the rays over the solid angle
subtended by an extended light source as seen from the point being
shaded.
34
Radiosity Methods

Disadvantage of Ray Tracing


Use of a directionless ambient-lighting term to account for all other
global lighting conditions.
For more accurate treatment of inter object reflections,
approaches based on thermal-engineering models for the
emission and reflection of radiation are used to eliminate the
need for the ambient-lighting term.



Assume the conservation of light energy in a closed environment.
All energy emitted or reflected by every surface is accounted for by its
reflection from or absorption by other surfaces.
The rate at which energy leaves a surface, called its RADIOSITY, is
the sum of the rate at which the surface emits energy and reflects or
transmits it from that surface or other surfaces.
35
Radiosity Methods


They first determine all the light interactions in an
environment in a view-independent way.
Then one or more views are rendered with only the
overhead of visible-surface determination and
interpolation shading.
36
Radiosity Methods

Radiosity methods allow any surface to emit light.
 All

light sources are modeled inherently as having area.
Break up the environment into a finite number n of
discrete patches.
 Each
patch emits and reflects light uniformly.
 It is an opaque Lambertian diffuse emitter and reflector.
 For surface patch i,
Bi,Bj : the Radiosities of patches i
Bi  E i   i

i j n
B j F j i
Aj
Ai
and j
Ei: the rate at which light is
emitted from patch i
ρi : patch i’s reflectivity
Fj-i: dimensionless form factor.
37
Radiosity Methods

The form factor specifies the fraction of energy
leaving the entirety of patch j that arrives at the
entirety of patch i, taking into account the shape and
relative orientation of both patches and the presence
of any obstructing patches.
Bi  E i   i

i j n

B j F j i
Aj
Ai,Aj : the areas of patches i and j
Ai
The above equation states that the energy leaving a
unit area of surface is the sum of the light emitted
plus the light reflected.
38
Radiosity Methods


A simple reciprocity relationship holds between form factors
in diffuse environments.
AF
 A F
Rearranging terms,
i
Bi  E i   i

i j
j
ji
B j Fi  j
i j n

The interaction of light among the patches in the environment
can be stated as a set of simultaneous equations.
39
Radiosity Methods

The above equation should be solved for each band
of wavelength considered in the lighting model.
 ρi

and Ei are wavelength dependent.
The form factors are independent of wavelength and
are solely a function of geometry.
40
Radiosity Methods

Computation of a form
factor
 The
form factor from
differential area dAi to
differential area dAj is
dF di  dj 
cos  i cos 
r
2
j
H ij dA j
41
Radiosity Methods

Computation of a form factor
 To
determine Fdi-j, the form factor from differential area
dAi to finite area Aj, we need to integrate over the area of
patch j.
Fdi  j 

cos  i cos 
r
Aj
2
j
H ij dA j
 The
form factor from Ai to Aj is the area average over
patch i:
Fi  j 
1
Ai

Ai A j
cos  i cos 
r
2
j
H ij dA j dA i
42
Radiosity Methods

Computation of a form factor
 Computing
Fdi-j is equivalent to projecting those parts of
Aj that are visible from dAi onto a unit hemisphere
centered about dAi, projecting this projected area
orthographically down onto the hemisphere’s unit circle
base and dividing by the area of the circle.
43
Radiosity Methods

Computation of a form factor
 Rather
than analytically projecting each Aj onto a
hemisphere, an efficient image-precision algorithm is
developed that projects onto the upper half of a cube
centered about dAi, with the cube’s top parallel to the
surface dAi.
44
Radiosity Methods

Computation of a form factor
45
Radiosity Methods

Computation of a form factor
46
Radiosity Methods
47