Principles of Computer Graphics

T e o / y and Practice Using OpenGL and /Maya@

Principles of Computer Graphics
Theow and Practice Using OpenGL and ~ a ~ a *

Shalini Govil-Pai
Sunnyvale, CA, U.S.A.

a- Springer
Shalini Govil-Pai
896 Savory Drive,
Sunnyvale, CA 94087

Email: sgovil @gmail.com

Library of Congress Cataloging-in-Publication Data

Govil-Pai, Shalini
Principles of Computer Graphics: Theory and Practice Using OpenGL and Maya@/ Shalini
Govil-Pai
p.cm.
Includes bibliographical references and index.

ISBN: 0-387-95504-6 (HC) e-ISBN 0-387-25479-X Printed on acid-free paper

ISBN-13: 978-0387-95504-9 e-ISBN-13: 978-0387-25479-1

O 2004 Springer Science+Business Media, Inc.

All rights resewed. This work may not be translated or copied in whole or in part without
the written permission of the publisher (Springer Science+Business Media, Inc., 233 Spring
Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or
scholarly analysis. Use in connection with any form of information storage and retrieval,
electronic adaptation, computer software, or by similar or dissimilar methodology now
know or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks and similar terms,
even if the are not identified as such, is not to be taken as an expression of opinion as to
whether or not they are subject to proprietary rights.

Printed in the United States of America.

Alias and Maya are registered trademarks of Alias Systems Corp. in the United States
andlor other countries.

9 8 7 6 5 4 3 2 1 SPIN 10879906 (HC) 1 11412601 (eBK)

springeronline.com
Contents
Preface vii

Section 1 1
1 From Pixels to Shapes 3
1.1 Complex Display Systems 4
1.2 Game Buffers 6
1.3 Coordinate Systems: How to identify pixel points 9
1.4 Shapes and Scan Converting 11

2 Making Them Move 27

2.1 Vectors and Matrices 28
2.2 2D Object Transformations 32
2.3 Homegenous Coordinates and Composition of
Matrix Transformations 41

3 Pixels, Images and Image Files 49

3.1 Raster Image Files 48
3.2 Bitmaps and Pixmaps 51
3.3 Computer Display Systems 56
3.4 Image Enhancements 60

4 Let The Games Begin 67

4.1 What is a Game? 68
4.2 Game Design 69
4.3 Implementing the Game 72

Section 2 81
5 3D Modeling 83
5.1 The 3D System 84
5.2 3D Modeling 90
5.3 3D Modeling Primitive Shapes 95
5.4 3D Modeling Generic Shapes 100
5.5 3D Transformation 104
5.6 Viewing in 3D 107
5.7 Hierarchical Modeling Using Transformations 118
vi
XV~ CONTENTS

6 Rendering, Shading and Lighting

6.1 What is Rendering
6.2 Hidden Surface Removal
6.3 Light Reflectance Model
6.4 CG: Reflectance Model
6.5 The Normal Vectors
6.6 Shading Models
6.7 Texture Mapping

7 Advanced Techniques
7.1 Advanced Modeling
7.2 Advanced Rendering Techniques

8 And Finally, Introducing Maya

8.1 Maya Basics
8.2 Modeling 3D Objects
8.3 Applying Surface Material
8.4 Composing the World
8.5 Lighting the Scene

Section 3
9 Animation
9.1 Traditional Animations
9.2 3D Computer Animation - Interpolations
9.3 The Principles of Animation
9.4 Advanced Animation Techniques

10 Viewpoint Animation
10.1 Animating the Camera inthe Snowy Animation
10.2 Building up a Real Time 3D Game

11 Lights, Camera, Action!

11.1 Pre-Production
11.2 Production
11.3 Post-Production
11.4 Finally, Our Movie!

Appendix A
Appendix B
Appendix C

Bibliography

Index of Terms
Preface
Computer Graphics: the term has become so widespread now, that we rarely stop
to think about what it means. What is Computer Graphics? Simply defined,
Computer Graphics (or CG) is the images generated or modified on a computer.
These images may be visualizations of real data or imaginary depictions of a
fantasy world.
The use of Computer Graphics effects in movies such as The Incredibles and
games such as Myst have dazzled millions of viewers worldwide. The success of
such endeavors is prompting more and more people to use the medium of
Computer Graphics to entertain, to educate, and to explore.
For doctors, CG provides a noninvasive way to probe the human body and to
research and discover new medications. For teachers, CG is an excellent tool to
visually depict concepts to their students. For business people, CG has come to
signify images of charts and graphs used for analysis of data. But for most of us,
CG translates into exciting video games, special effects and entire films-what are
often referred to as CG productions. This entertainment aspect of CG is what has
made it such a glamorous and sought-after field.
Ten years ago, CG was limited to high-end workstations, available only to an
elite few. Now, with the advances in PC processing power and the availability of
3D graphics cards, even high school students can work on their home PC to
create professional quality productions
The goal of this book is to expose you to the fundamental principles behind
modern computer graphics. We present these principles in a fun and simple
manner. We firmly believe that you don't have to be a math whiz or a high tech
computer programmer to understand CG. A basic knowledge of trigonometry,
algebra, and computer programming is more than sufficient.
As you read this book, you will learn the bits and bytes of how to transform
your ideas into stunning visual imagery. We will walk you through the processes
that professionals employ to create their productions, Based on the principles
that we discuss, you will follow these processes step and step, to design and
create your own games and animated movies.
We will introduce you to the OpenGL API—a graphics library that has
become the de facto standard on all desktops. We will also introduce you to the
workings of Maya, a 3D software package. We will demonstrate the workings of
the Maya Personal Learning Edition—a (free) download is required.
viii PREFACE

Organization of the Book

The book is organized into three sections. Every section has detailed OpenGL
code and examples. Appendix B details how to install these examples on your
desktop.

Section 1: The Basics

The first section introduces the most basic graphics principles. In Chapter 1, we
discuss how the computer represents color and images. We discuss how to
describe a two-dimensional (2D) world, and the objects that reside in this world.
Moving objects in a 2D world involves 2D transformations. Chapter 2 describes
the principles behind transformations and how they are used within the CG
world. Chapter 3 discusses how the computer saves images and the algorithms
used to manipulate these images. Finally, in Chapter 4, we combine all the
knowledge from the previous chapters to create our very own version of an
arcade game.

Section 2: It’s 3D time

Section 2 will expand your horizon from the 2D world to the 3D world. The 3D
world can be described very simply as an extension of the 2D world. In Chapter
5, we will introduce you to 3D modeling. Chapter 6 will discuss rendering: you
will have the opportunity to render your models from Chapter 5 to create
stunning visual effects. Chapter 7 is an advanced chapter for those interested in
more advance concepts of CG. We will introduce the concept of Nurbs as used
in modeling surfaces. We will also introduce you to advanced shading concepts
such as ray tracing. Chapter 8 focuses on teaching the basics of Maya and the
Maya Personal Learning Edition of Maya (Maya PLE). Maya is the most popular
software in the CG industry and is extensively used in every aspect of
production. Learning the basics of this package will be an invaluable tool for
those interested in pursuing this area further.

Section 3: Making Them Move

Section 3 discusses the principles of animation and how to deploy them on the
computer. In Chapter 9, we discuss the basic animation techniques. Chapter 10
discusses a mode of animation commonly deployed in games, namely, viewpoint
animation. In Chapter 11, you will have the opportunity to combine the working
knowledge from the previous chapters to create your own movie using Maya.

Appendices
In Appendix A, you will find detailed instructions on how to install the OpenGL
and GLUT libraries. Appendix B describes how to download, install the sample
code that is detailed in this book. You will also find details on how to compile
PREFACE ix

and link your code using the OpenGL libraries. Appendix C describes the Maya
PLE and how to download it.

OpenGL and Maya

Every concept discussed in the book is followed by examples and exercises

using C and the OpenGL API. We also make heavy use of the GLUT library,
which is a cross-platform OpenGL utility toolkit. The examples will enable you
to visually see and appreciate the theory explained. We do not expect you to
know OpenGL, but we do expect basic knowledge in C and C++ and knowledge
of compiling and running these programs. Some chapters detail the workings of
Maya, a popular 3D software package. Understanding Maya will enable you to
appreciate the power of the CG concepts that we learn in the book.

Why are we using OpenGL and GLUT?

OpenGL is now a widely accepted industry standard and is used by many (if not
all) professional production houses. It is not a programming language but an
API. That is, it provides a library of graphics functions for you to use within your
programming environment. It provides all the necessary communication
between your software and the graphics hardware on your system.
GLUT is a utility library for cross-platform programming. Although our code
has been written for the Windows platform, GLUT makes it easier to compile the
example code on other platforms such as Linux or Mac. GLUT also eliminates
the need to understand basic Windows programming so that we can focus on
graphics issues only.

Why are we using Maya?

Some concepts in the book will be further illustrated with the help of industry
leading 3D software Maya. Academy-Award winning Maya 3D animation and
effects software has been inspired by the film and video artists, computer game
developers, and design professionals who use it daily to create engaging digital
imagery, animation, and visual effects. Maya is used in almost every production
house now, so learning the basics of it will prove to be extremely useful for any
CG enthusiast. In addition, the good folks at Alias now let you download a free
version of Maya (Maya PLE) to use for learning purposes.

The system requirements for running the examples in this book, as well as for
running Maya PLE are as follows:

Software Requirements
• Windows 2000 or higher
•C/C++ compiler such as Microsoft Visual Studio on Windows or GCC
(Gnu Compiler Collection) on Unix
x PREFACE

Hardware Requirements
• Intel Pentium II or higher/AMD Athlon processor
• 512 MB RAM
• Hardware-accelerated graphics card (comes standard on most systems)

In addition, we expect some kind of Internet connectivity so that you can

download required software.

Intended Audience
This book is aimed at undergraduate students who wish to gain an overview of
Computer Graphics. The book can be used as a text or as a course supplement
for a basic Computer Graphics course.
The book can also serve as an introductory book for hobbyists who would
like to know more about the exciting field of Computer Graphics, and to help
them decide if they would like to pursue a career in it.

Acknowledgments

The support needed to write and produce a book like this is immense. I would
like to acknowledge several people who have helped turn this idea into a reality,
and supported me through the making of it:
First, my husband, Rajesh Pai, who supported me through thick and thin. You
have been simply awesome and I couldn't have done it without your constant
encouragement.
A big thanks to my parents, Anuradha and Girjesh Govil, who taught me to
believe in myself, and constantly egged me on to publish the book.
Thanks to Carmela Bourassa of Alias software, who helped provide
everything I needed to make Maya come alive.
A very special thanks to my editor, Wayne Wheeler, who bore with me
through the making of this book and to the entire Springer staff who helped to
produce this book in its final form.
I would like to dedicate this book to my kids, Sonal and Ronak Pai, who
constantly remind me that there is more to life than CG.
PREFACE xi

CG technology is emerging and changing every day. For example, these days,
sub-division surfaces, radiosity, and vertex shaders are in vogue. We cannot
hope to cover every technology in this book. The aim of the book is to empower
you with the basics of CG-providing the stepping-stone to pick up on any CG
concept that comes your way.
A key tenet of this book is that computer graphics is fun. Learning about it
should be fun too. In the past 30 years, CG has become pervasive in every aspect
of our lives. The time to get acquainted with it is now—so read on!
Section I
The Basics
Imagine how the world would be if computers had no way of drawing pictures
on the screen. The entire field of Computer Graphics-flight simulators, CAD
systems, video games, 3D movies-would be unavailable. Computers would be
pretty much what they were in the 1960s - just processing machines with
monitors displaying text in their ghostly green displays.
Today, computers do draw pictures. It's important to understand how
computers actually store and draw graphic images. The process is very different
from the way people do it. First, there's the problem of getting the image on the
screen. A computer screen contains thousands of little dots of light called pixels.
To display a picture, the computer must be able to control the color of each pixel.
Second, the computer needs to know how to organize the pixels into meaningful
shapes and images. If we want to draw a line or circle on the screen, how do we
get the computer to do this?
The answers to these questions form the basis for this section. You will learn
how numbers written in the frame buffer control the colors of the pixels on the
screen. We will expose you to the concept of two-dimensional coordinate
systems and how 2D shapes and objects can be drawn and transformed in this 2D
world.You will learn the popular algorithms used to draw basic shapes such as
lines and circles on the computer.
These days, three-dimensional graphics is in vogue. As a reader, you too
must be eager to get on to creating gee-whiz effects using these same principles.
It is important, however, to realize that all 3D graphics principles are actually
extensions of their 2D counterparts. Understanding concepts in a 2D world is
much easier and is the best place to begin your learning. Once you have mastered
2D concepts, you will be able to move on to the 3D world easily. At every step,
you will also have the opportunity to implement the theory discussed by using
OpenGL.
At the end of the section, we shall put together everything we have learned
to develop a computer game seen in many video arcades today.
Chapter *I
From Pixels to Shapes

The fundamental building block of all computer images is the picture element,
or the pixel. A pixel is a dot of light on the computer screen that can be set to
different colors. An image displayed on the computer, no matter how complex,
is always composed of rows and columns of these pixels, each set to the
appropriate color and intensity. The trick is to get the right colors in the right
places.
Since computer graphics is all about creating images, it is fitting that we
begin our journey into the computer graphics arena by first understanding the
pixel. In this chapter, we will see how the computer represents and sets pixel
colors and how this information is finally displayed onto the computer screen.
Armed with this knowledge, we will explore the core graphics algorithms used
to draw basic shapes such as lines and circles.
In this chapter, you will learn the following concepts:
What pixels are
w How the computer represents color
How the computer displays images
w The core algorithm used to draw lines and circles
w How to use OpenGL to draw shapes and objects
1.1 Computer Display Systems
The computer display, or the monitor, is the most important device on the
computer. It provides visual output from the computer to the user. In the
Computer Graphics context, the display is everything. Most current personal
computers and workstations use Cathode Ray Tube (CRT) technology for their
displays.
As shown in Fig. 1.1, a CRT consists of
An electron gun that emits a beam of electrons (cathode rays)
A deflection and focusing system that directs a focused beam of
electrons towards specified positions on a phosphorus-coated screen
A phosphor-coated screen that emits a small spot of light proportional
to the intensity of the beam that hits it
The light emitted from the screen is what you see on your monitor.

Deflection and focusing system

Phosphor
coated screen

Fig.l.1: A cathode ray tube

The point that can be lit up by the electron beam is called a pixel. The
intensity of light emitted at each pixel can be changed by varying the number of
electrons hitting the screen. A higher number of electrons hitting the screen will
result in a brighter color at the specified pixel. A grayscale monitor has just one
phosphor for every pixel. The color of the pixel can be set to black (no electrons
hitting the phosphor), to white (a maximum number of electrons hitting the
phosphor), or to any gray range in between. A higher number of electrons hitting
the phosphor results in a whiter-colored pixel.
A color CRT monitor has three different colored phosphors for each pixel.
Each pixel has red, green, and blue-colored phosphors arranged in a triangular
group. There are three electron guns, each of which generates an electron beam
to excite one of the phosphor dots, as shown in Fig.l.2. Depending on the
monitor manufacturer, the pixels themselves may be round dots or small squares,
as shown in Fig. 1.3.
 Electron Guns

el composed of a triad

Fig.l.2: Color CRT uses red green and blue triads

Because the dots are close together, the human eye bses the three red, green, and
blue dots of varying brightness into a single dothquare that appears to be the
color combination of the three colors. (For those of us who missed art class in
school, all colors perceived by humans can be formed by the right brightness
combination of red, green, and blue color.)
Conceptually, we can think of the screen as a discrete two-dimensional array (a
matrix) of pixels representing the actual layout of the screen, as shown in Fig. 1.3.
The number of rows and columns of pixels that can be shown on the screen
is called the screen resolution. On a display device with a resolution of 1024 x
768, there are 768 rows (scan lines), and in each scan line there are 1024 pixels.
That means the display has 768 x 1024=786,432 pixels! That is a lot of pixels
packed together on your 14-inch monitor. Higher-end workstations can achieve
even higher resolutions.
Fig.l.4 shows two images displayed in different resolutions. At lower
resolutions, where pixels are big and not so closely packed, you can start to
notice the "pixelated" quality of the image as in the image shown on the right.
At higher resolutions, where pixels are packed close together, your eye perceives
a smooth image. This is why the resolution of the display (and correspondingly
that of the image) is such a big deal.
You may have heard the term dpi, which stands for dots per inch. The word
dot is really referring to a pixel. The higher the number of dots per inch of the

Fig.l.3: Computer display: rows and columns of pixels

 screenlimage, the higher the resolution and hence the crisper the image.
We have seen we can represent a computer display as a matrix of pixels. But
how can we identify an individual pixel and set its color? And how can we then
organize the pixels to form meaningfid images? In the next section, we explore
how pixel colors are set and manipulated.

Fig.l.4: The same image at different reolsutions

1.2 Frame Buffers

The light on the screen generated by the beam of electrons in our CRT fades
quickly-in 10 to 60 microseconds. In order to keep a picture on the screen for
a while, the picture needs be redrawn before it disappears off the screen. This is
called refreshing the screen. Most display systems use raster scan technology to
perform the refresh process. In this technology, the electron beam is directed
discretely across the screen, one row at a time from left to right, starting at the
upper left corner of the screen. When the beam reaches the bottommost row, the
process is repeated, effectively refreshing the screen.
Raster scan systems use a memory buffer called frame buffer (or refresh
buffer) in which the intensities of the pixels are stored. Refreshing the screen is
performed using the information stored in the frame buffer. You can think of
frame buffer as a two dimensional array. Each element of the array keeps the

frame buffer pixels in the display

Fig.l.5: Monochrome display: frame buffer for turning pixels on and off
intensity of the pixel on the screen corresponding to that element.
For a monochrome display, the frame buffer has one bit for each pixel. The
display controller keeps reading from the frame buffer and turns on the electron
gun only if the bit in the buffer is as shown in Fig. 1.5.
Systems can have multiple buffers. Foreground buffers draw directly into the
window specified. Sometimes a background buffer is also used. The background
buffer is not displayed on the screen immediately. We shall talk about buffering
modes in more detail when we study animation.

How about color?

You may recall from school physics that all colors in the world can be repre-
sented by mixing differing amounts of the three primary colors, namely, red,
green, and blue. In CG, we represent color as a triplet of the Red, Green, and
Blue components. The triplet defines the final color and intensity. This is called
the RGB color model. Color Plate 1 shows an image of Red, Green and Blue
circles and the resultant colors when they intersect.
Some people use a minimum of 0 and a maximum of 255 to represent the
intensities of the three primaries, and some people use a floating-point number
between 0 and 1. In this book (as is the case in OpenGL), we shall use 0 to
represent no color and 1.0 to represent the color set to its maximum intensity.
Varying the values in the RGB triplet yields a new color. Table 1.1 lists the RGB
components of common colors.
On color systems, each pixel element in the frame buffer is represented by an
RGB triplet. This triplet controls the intensity of the electron gun for each of the
red, green, and blue phosphors, respectively of the actual pixel on the screen.
Our eye perceives the final pixel color to be the color combination of the three
colors.
Each pixel color can be set independent of the other pixels. The total number
of colors that can be displayed on the screen at one time, however, is limited by
the number of bits used to represent color. The number of bits used is called the
color resolution of the monitor.
For lower resolution systems like VGA monitors, the color resolution is

Table 1.1: The RGB components of common colors

usually 8 bits. Eight-bit systems can represent up to 256 colors at any given time.
These kinds of systems maintain a color table. Applications use an index (from
1 to 256) into this color table to define the color of the screen pixel. This mode
of setting colors is called color index mode and is shown in Fig. 1.6.
Of course, if we change a color in this table, any application that indexes into
this table will have its color changed automatically. Not always a desirable
effect!
Most modern systems have a 24-bit color resolution or higher. A 24-bit
system (8 bits for the red channel, 8 for the green channel, and 8 for the blue
channel) can display 16 million colors at once. Sometimes an additional 8 bits is
added, called the alpha channel. We shall look into this alpha channel and its
uses later when we learn about fog and blending.
With so many colors available at any given time, there is no need for a color
table. The colors can be referred to directly by their RGB components. This way

INDEX

\
Index value Final Pixel color
displayed

Color Table

Fig. 1.6:Color index mode

of referring to colors is called RGB mode. We shall employ the RGB mode to
refer to color throughout the rest of this book.
We have seen how pixel colors are stored and displayed on the screen. But
we still need to be able to identify each pixel in order to to set its color. In the
next section, we shall see how to identify individual pixel points that we want to
paint.
1.3 Coordinate Systems: How to IdentifL Pixel Points

Coordinates are sets of numbers that describe position-position along a line, a

surface of a sphere, etc. The most common coordinate system for plotting both
two dimensional and three-dimensional data is the Cartesian coordinate system.
Let us see how to use this system to identify point positions in 2D space.
The Cartesian coordinate system is based on a set of two straight lines called
the axes. The axes are perpendicular to each other and meet at the origin. Each
axis is marked with the distances from the origin. Usually an arrow on the axis
indicates positive direction. Most commonly, the horizontal axis is called the x-
axis, and the vertical axis is called the y-axis.
Fig.l.7 shows a Cartesian coordinate system with an x- and a y-axis. To
define any point P in this system, we draw two lines parallel to the xy-axes. The

Fig.l.7: Cartesian coordinate system

values of x andy at the intersections completely define the position of this point.
In the Cartesian coordinate system, we label this point as (x,y). x and y are called
the coordinates of the point (and could have a negative value). In this system, the
origin is represented as (0,0), since it is at 0 distance from itself. A coordinate
system can be attached to any space within which points need to be located.
Coming back to the world of computers, recall that our computer display is
represented physically in terms of a grid of pixels. This grid of pixels can be
defined within its own Cartesian coordinate system. Typically, it is defined with
an origin at the upper left corner of the screen. We refer to this Cartesian space
as the physical coordinate system. Within this system, each pixel can then be
uniquely identified by its (x,y) coordinates, as shown in Fig.l.8.
 Fig.l.8: Identifying pixels on the screen
Now, consider an applicatibn window being shown on this display. We can
specify a Cartesian space within which the application resides within. In Fig. 1.9,
the x- coordinates of the window define a boundary ranging from -80 to 80 and
the y-coordinates fiom -60 to 60. This region is called the clipping area, and is
also referred to as the logical or world coordinate system for our application.
This is the coordinate system used by the application to plot points, draw lines

Fig.l.9: Application (clipping) area

and shapes, etc. Objects within the clipping area are drawn, and thosc outside
this area are removed or clipped from the scene. The clipping area is mapped
onto a physical region in the computer display by mapping the application
boundaries to the physical pixel boundaries of the window.
If the clipping area defined matches the (physical) resolution of thc window,
then each call to draw an (x,y)point (with integer values) in the world coordinate
system will have a one-to-onc mapping with a corresponding pixel in the
physical coordinate system.
For most applications, the clipping area does not match the physical size of
the window. In this case, the graphics package needs to perform a transformation
 Clipping area:
320 by 240 units
(world coordinates)

Fig.l.10: Mapping Clipping Area onto the window

fiom the world coordinate system being used by the application to the physical
window coordinates. This transformation is determined by the clipping area, the
physical size of the window, and another setting known as the viewport. The viewport
defines the area of the window that is actually being used by the application.
For now, we will assume that the viewport is defined as the entire window
(but this is not always necessary, as shown in Fig. 1.11).
-,. --

~ " " ' - ~ i n d o warea:

r~ 320 bv 240 pixels

0
0
0
0
8

clipping area:
160 by 120
(in world coordinates)
Fig.l.11: Viewport of a Window

Example Time

Let us run our first OpenGL program to get a handle on some of the concepts we
have just learned. For information on OpenGL and GLUT and how to install it
on your system refer to Appendix A on details. For information on how to
download the sample example code from the Internet and how to compile and
link your programs, refer to Appendix B.
The following example displays a window with physical dimensions of 320
by 240 pixels and a background color of red. We set the clipping area (or the
world coordinates) to start from (0,O) and extend to (160,120). The viewport is
set to occupy the entire window, or 320 by 240 pixels. This setting means that
every increment of one coordinate in our application will be mapped to two pixel
increments in the physical coordinate system (application window boundaries
(0,O) to (160,120) vs. physical window boundaries (0,O) to (320,240)). If the
viewport had been defined to be only 160 by 120 pixels, then there would have
been a one-to-one mapping from points in the world coordinate space to the
physical pixels.
However, only one fourth of the window would have been occupied by the
application! Depending on where you install the example files, you can find the
source code for this example in: Examplel-l/Examplel-1. cpp.

IIExamplel-I .cpp: A simple example to open a window

I1the windows include file, required by all windows apps
Anclude <windows.h>

I1the glut file for windows operations

11it also includes g1.h and g1u.h for the OpenGL library calls
Anclude <gl\glut.h>

void Displaylvoid)
{
llclear all pixels with the specified clear color
glClear(GL-COLOR-BUFFER BIT);
Ildonlt wait, start flushing OpenGL calls~odisplay buffer
glFlush0;
1
void initlvoidl{
llset the clear color to be red
glClearColor(1.O,O.OIO.O,l -0);
llset the viewport to be 320 by 240, the initial sii of the window
glViewport(0,0,320,240~;
11set the 2D clipping area
gluOrtho2D(O.O, 160.0,0.0, 120.01;
1
void mairdint argc, char* argv01
{
glutlnitDisplayMode(GLUT-SINGLE I GLUT-RGB);
 glutlnitWindowSi 1320,240);
glutCreateWindowVMy first OpenGL Window");
init0;
glutDisplayFunclDisplay1;
glutMainLoop0;
1
Since this is our first OpenGL program, Let us understand the example line by
line.

The Include Files

There are only two include files:

Anclude cwindows.h>
Anclude cgl\glut.h>

The wind0ws.h is required by all windows applications. The header file g1ut.h
includes the GLUT library functions as well as g1.h and glu.h, the header files for the
OpenGL library functions. All calls to the glut library functions are prefixed with
glut. Similarly, all calls to the OpenGL library functions start with the prefix gl or glu.

The Body
Let us look at the main program first. The line

glutlnitDisplayMode1GLUT-SINGLE I GLUT-RGB);
tells the GLUT library what type of display mode to use when creating the
application window. In this case, we specify that we will be using only a single
frame buffer (GLUT-SINGLE) and we want to specify colors in the RGB color
mode (GLUT-RGB). We will discuss more about frame buffers in a later
chapter.
The next call

initializes the window to have an initial size (physical resolution) of 320 by 240
pixels.
The call

glutCreateWindowVMy First OpenGL Window");

actually creates the window with the caption "My First OpenGL Window".
The next function

initializes some of the OpenGL parameters before we actually display the

rendered window.
OpenGL and glut work with the help of callback functions. Events that occur
on the computer (such as mouse clicks, keyboard clicks, moving the window
etc.) that you wish your program to react to need to be registered with OpenGL
as callback functions. When the event occurs, OpenGL automatically calls the
function registered to react to the event appropriately.
The function

registers the callback function for display redrawing to be the function Display.
The Display callback is triggered whenever the window needs to be redrawn, and
GLUT will automatically call the Display function for you. When does the
window need to be redrawn? When you first display the window, when you
resize the window, or even when you move the window around. We shall see
what the init function and the Display function actually do in a bit.
Finally, we make a call to the glut library function

This function simply loops, monitoring user actions and making the necessary
calls to the specified callback functions (in this case, the 'Display' function) until
the program is terminated.

The init(, function

The init function itself is defined to initialize the GL environment. It does this by
making three calls:

This gl library command sets the color for clearing out the contents in the h m e buffer
(which then get drawn into the window). It expects the RGB values, in that order, as
parameter,^ as well as the alpha component of the color. For now, we set the alpha to
always be 1.The above command will set the clear color to be pure red. Try experimenting
with different clear colors and see what effect this has on the window display.
Next, we define the viewport to be equal to the initial size of the window by
calling the function
 And we set the clipping area, or our world coordinate system, to be (0,O) to
(160,120) with the glu library command

The Display(' function

The Display function simply makes two OpenGL calls:

On a computer, the memory (frame buffer) holding the picture is usually filled
with the last picture you drew, so you typically need to clear it with some
background color before you start to draw the new scene.
OpenGL provides glClear as a special command to clear a window. This
command can be much more efficient than a general-purpose drawing command
since it clears the entire frame buffer to the current clearing color. In the present
example, we have set the clear color earlier to be red.
In OpenGL, the frame buffer can be further broken down into buffers that
hold specialized information.
The color buffer (defined as GL-COLORBUFFERBIT) holds the color
information for the pixel. Later on, we shall see how the depth buffer holds depth
information for each pixel.
The single parameter to glClear() indicates which buffers are to be cleared.
In this case, the program clears only the color buffer.
Finally, the hnction

forces all previously issued OpenGL commands to begin execution. If you are
writing your program to execute within a single machine, and all commands are
truly executed immediately on the server, glFlush() might have no effect.
However, if you're writing a program that you want to work properly both with
and without a network, include a call to @Flush() at the end of each frame or
scene. Note that glFlush() doesn't wait for the drawing to complete -it just
forces the drawing to begin execution.
Voila: when you run the program you will see a red window with the caption
"My First OpenGL window". The program may not seem very interesting, but it
demonstrates the basics of getting a window up and running using OpenGL.
Now that we know how to open a window, we are ready to start drawing into it.

Plotting Points
Objects and scenes that you create in computer graphics usually consist of a
combination of shapes arranged in unique combinations. Basic shapes, such as
points, lincs, circles, etc., are known as graphics primitives. The most basic
primitive shape is a point.
In the previous section, we considered Cartesian coordinates and how we can
map the world coordinate system to actual physical scrccn coordinatcs. Any
point that we define in our world has to be mapped onto the actual physical
screen coordinates in order for the correct pixel to light up. Luckily, for us,
OpenGL handles all this mapping for us; we just have to work within the extent
of our defincd world coordinate systcm. A point is reprcscntcd in OpcnGL by a
set of floating-point numbers and is called a vertex.
Wc can draw a point in our window by making a call to the gl library function

The {2,3,4) option indicates how many coordinates define the vertex and the
{s,i,d,f) option defines whether the arguments are short, integers, double
precision, or floating-point values. By default, all integer values arc internally
converted to floating-point values.
For cxamplc, a call to

refers to a vcrtcx point (in world coordinate spacc) at coordinates (1.0,2.0).

Almost all library functions in OpenGL use this format. Unless otherwise stated,
wc will always usc the floating-point vcrsion of all functions. To tcll OpcnGL
what set of primitives you want to define with the vertices, you bracket each set
of vertices between a call to

glBegin0 and glEnd0.

The argument passed to glBegin() detcrmines what sort of geometric primitive it

is. To draw vertex points, the primitive used is GL-POINTS. We modify
Examplel-l to draw four points. Each point is at a distance of (10,lO)
coordinates away from the corners of the window and is drawn with a different
color. Compile and execute the code shown below. You can also find the code for
this example under Exarnplel_2/Example 1-2.cpp..

I1 Examplel-2.cpp: let the drawing begin

#include cwindows.h>
Anclude <gRglut.h>

void Display(void1
 llclear all piiels with the specified clear color
glClear(GL-COLOR-BUFFER-BIT);
lldraw the four points in four colors
glBeginlGL-POINTS);
glColor3f(O.O, 1.O, 0.0); I1green
gNertex2fll O.,10.1;
glColor3f(l.Or 1.0,O.O); I1yellow
glVertex2f(lO.,ll0.);
glColor3fl0.0, 0.0, 1.Dl; I1blue
gNertex2fU 50.,110.1;
glColor3fU.Or 1.0, 1.0); I1white
glVertex2f~l50.,10.);
glEnd0;

lldont wait, start flushing OpenGL calls to display buffer

glFlush0;
I
void reshape (int w, int h)
{
11on reshape and on startup, keep the viewport to be the entire size of the window
glviewport (O,O, (GLsiiei) w, (GLsizeil hl;
glMatrixMode IGL-PROJECTION);
glloadldentity 0;
I1keep our world coordinate system constant
gluOrtho2D10.0,160.0,0.0,120.01;
I
void initlvoidl{
glClearColor(l.0,0.0,0.0,1.0);
I1set the point size to be 5.0 pixels
glPointSize(5.01;
I
void mainlint argc, char* argvUl
{
glutlnitDisplayMode(GLUT-SINGLE I GLUT-RGB);
glutlnitWindowSie 1320, 240);
glutCreateWindowl'My First OpenGL Window");
init();
glutDisplayFunc(Display);
glutReshapeFunc(reshape1;
glutMainLoopl1;
I
Most of the code should be self-explanatory.
The main function sets up the initial OpenGL environment.
In the init() function, we set the point size of each vertex drawn to be 5 pixels,
by calling the function

The parameter defines the size in pixels of the points being drawn. A 5 pixel
point is large enough for us to see it without squinting our eyes too much!
In this function, we also we set the clear color to be red. The Display function
defines all the drawing routines needed
The function call

glColor3fl0.0, 1.Of 0.0); 11 green

sets the color for the next openGL call. The parameters are, in order, the red,
green, and blue components of the color. In this case, we redefine the color
before plotting every vertex point. We define the actual vertex points by calling
the function

with the appropriate (x,y) coordinates of the point. In this example, we define
two callback functions. We saw how to define the callback function for
redrawing the window. For the rest of the book, we will stick with the convention
of this function being called "Display".
In this example we define the viewport and clipping area settings in a new
callback function called reshape. We register this callback function with
OpenGL with the command

This means that the reshape function will be called whenever the window resizes
itself (which includes the first time it is drawn on the screen!) The function
receives the width and height of the newly shaped window as its arguments.
Every time the window is resized, we reset the viewport so as to always cover
the entire window. We always define the world coordinate system to remain
constant at ((0,0),(160,120)). As you resize the window, you will see that the
points retain their distance from the corners. What mapping is being defined? If
we change the clipping area to be defined as

you would see that the points maintain their distance from each other and not
from the corners of the window. Why?
1.4 Shapes and Scan Converting

We are all familiar with basic shapes such as lines and polygons. They are easy
enough to visualize and represent on paper. But how do we draw them on the
computer? The trick is in finding the right pixels to turn on!
The process by which an idealized shape, such as a line or a circle, is
transformed into the correct "on" values for a group of pixels on the computer is
called scan conversion and is also referred to as rasterizing.
Over the years, several algorithms have been devised to make the process of
scan converting basic geometric entities such as lines and circles simple and fast.
The most popular line-drawing algorithm is the midpoint-line algorithm.
This algorithm takes the x- and y- coordinates of a line's endpoints as input and
then calculates the x,y-coordinate pairs of all the pixels in between. The algo-
rithm begins by first calculating the physical pixels for each endpoint. An ideal
line is then drawn connecting the end pixels and is used as a reference to
determine which pixels to light up along the way. Pixels that lie less than 0.5
units from the line are turned on, resulting in the pixel illumination as shown in
Fig.1 .l2.

Fig.l.12: Midpoint algorithm for line drawing

All graphic packages (OpenGL included) incorporate predefined algorithms

to calculate the pixel illuminations for drawing lines.
Basic linear shapes such as triangles and polygons can be defined by a series
of lines. A polygon is defined by n number of vertices connected by lines, where
n is the number of sides in the polygon. A quadrilateral, which is a special case
of a polygon is defined by four vertices, and a triangle is a polygon with three
vertices as shown in Fig. 1.13.
To specify the vertices of these shapes in OpenGL, we use the function that
we saw earlier:

To tell OpenGL what shape you want to create with the specified vertices, you
bracket each set of vertices between a call to glBegin() and a call to glEnd(). The
1) A line is defined by 2 vertices 2) A triangle is defined by 3 vertices

3) A quadrilateral is defined by 4 verticles 4) An n-polygon is defined by n vertices

Fig.l.13: Vertices needed to define different kinds of basic shapes

argument passed to glBegin() determines what sort of geometric primitive; some

of the commonly used ones are described in Table 1.2.
Primitive definition Meaning
GL-POINTS individual points
GL-LINES - -- ----- -- - --- - pair-of-vertices defining-a-line
-

GL-LINE-STRIP
- - ---- series-of-connected-lines
p
GL-TRIANGLES --- ----
r
strip---
of linked--triangles-- -
- - -

GL-POLY
- -
GON -- - -
vertices- define--a simple convex
-- -- polygon__I
1

Table 1.2: OpenGL geometric primitive types

Note that primitives are all straight-line primitives. There are algorithms like the
midpoint algorithm that can scan convert shapes like circles and other hyperbolic
fig.s. The basic mechanics for these algorithms are the same as for lines: figure
out the pixels along the path of the shape, and turn the appropriate pixels on.
Interestingly, we can also draw a curved segment by approximating its shape
using line segments. The smaller the segments, the closer the approximation.
For example, consider a circle. Recall from trigonometry that any point on a
circle of radius r (and centered at the origin) has an x,y-coordinate pair that can
be represented as a function of the angle theta the point makes with the axes, as
shown in Fig. 1.14.

P ( q =((r COS~),(r sine))

 Fig.l.14: Points along a circle

As we vary theta from 0 to 360 degrees (one whole circle), we can get the
(x,y) coordinates of points along the circle. So if we can just plot "enough" of
these points and draw line segments between them, we should get a fig. that
looks close enough to a circle. A sample code snippet to draw a circle with
approximately 100 points is shown below. Note that we need to add the center of
the circle to our equations to position our circle appropriately.

Wefine PI 3.1 415926535898

I1 cos and sin functions require angles in radians
-
11 recall that 2PI radians 360 degrees, a full circle

-
GLint circlegoints 100;
void MyCircle2flGLfloat centerx, GLfloat centery, GLfloat radius){
GLint i;
GLdouble theta;
glBeginIGL-POLYGON);
-
for (i 0;i < circlegoints; i+ + 1 {
theta- 2"Pl"ilcirclegoints; 11 angle in radians
glVertex2f(centerx+ radius"cosltheta1,
centery + radius"sin(theta1l;

Remember that the math functions cos and sin require angles in radians and
that 2PI radians make 360 degrees-hence our conversions in the code. We can
construct more complex shapes by putting these basic primitives together.
Shown below is a snippet of code to draw a stick figure of a person as shown in
 Fig.1.I5 Stick Figure
the Fig. 1.15. The figure is composed of lines, polygons, points and a circle. The
entire code can be found in Examplel-3/Examplel-3.cpp.

I1Example 1-3.cpp

void Displaylvoid)
{
llclear all pixels with the specified clear color
glClearlGL-COLOR-BUFFER-BIT);
glColor3fll.0,0.8,0.1 I;
MyCircle2f180.,85., 10.);

I1the eyes are black points

I1set the point size to be 3.0 pixels
glBeginlGL-POINTS);
glColor3fl0.0, 0.0, 0.0);
glVertex2R77.,88.1;
glVertex2ff83.,88.l;
glEndO;

I1polygonal body
glC0lor3fl0.8~0.0~0.9);
glBeginlGL-POLYGON);
glVertex2fl75.,75.);
gNertex2ff85.,75.);
glVertex2R100.,30.1;
gNertex2f160.,30.1;
glEndO;

llrectangular legs
glColor3fll.0,0.8,0.1);
glRectfl70.,5.,75.,30.);
glRectf185.,5.,90.,30.1;
 I1 but lines for hands!
glBeginIGL-LINES);
glVertex2f 174.,70.1; gNertex2f 150.,50.);
glEnd0;
glBeginlGL-LINES);
glVertex2f (86.,70.1; glVertex2f (110.,50.1;
glEnd0;

Ildon't wait, start flushing OpenGL calls to display buffer

glFlush0;
1

Note that with OpenGL, the description of the shape of an object being drawn
is independent of the description of its color. Whenever a particular geometric
object is drawn, it is drawn using the currently specified coloring scheme. Until
the color or coloring scheme is changed, all objects are drawn in the current
coloring scheme. Similarly, until the point or line sizes are changed, all such
primitives will be drawn using the most currently specified size. Try composing
your own objects by putting smaller primitives together.
If you run the above program, you may notice that the slanting lines appear
to be jagged. This is an artifact caused due to the algorithms that we employ to
rasterize the shapes and is known as aliasing.

An ti-Aliasing

The problem with most of the scan conversion routines is that the conversion is
jagged. This effect is an unfortunate consequence of our all or nothing approach
to illuminating pixels. At high resolutions, where pixels are close together, thi
effect is not noticeable, but on low-resolution monitors, it produces a harsh,
jagged look. Aliasing can be a huge problem in computer generated movies,
when you can sometimes actually see jagged lines crawling from scene to scene,
creating a disturbing effect.
A solution to this problem is called anti-aliasing. It employs the principle that
if pixels are set to different intensities, and if adjoining pixel intensities can be
properly manipulated, then the pixels will blend to form a smooth image. So
going back to the midpoint algorithm, as shown in Fig. 1.16, anti-aliasing would
turn pixels on with varying intensities (depending on how a one-unit thick line
would intersect with the pixels), instead of merely turning pixels on and off. This
process tends to make the image look blurry but more continuous.
1) An ideal line using end pixels 2) Illuminating the pixels 3) Exaggerated view
of jaggp line

1) A one pixel thick line 2) Setting at different intensities 3) Smooth

Fig.1.l6: Anti-aliasing a line

To deploy anti-aliasing in OpenGL, there are two steps we need to take:
1. To antialias points or lines, you need to turn on antialiasing with
glEnable(), passing in GL-POINT-SMOOTH or
GL-LINE-SMOOTH, as appropriate.
2. We also need to enable blending by using a blending factor. The
blending factors you most likely want to use are GL-SRC-ALPHA
(source) and GL-ONE-MINUS-SRC-ALPHA (destination).
To anti-alias our stick figure, we add a few more calls in the init function
from Examplel-3, as shown in the code below:

glEnable (GL-LINE-SMOOTH);
glEnable (GL-BLEND);
glBlendFunc(GL-SRC-ALPHA,GL-ONE-MINUS-SRC-ALPHA);
glHintlGL-LINE-SMOOTH-HINT I GL-POLYGON-SMOOTH-HINT, GL-DONT-CARE);

You can find the entire source code in Examplel-4/Examplel-4.cpp.When

you run this example, look at the hands carefully. You will notice they seem
smoother than from Examplel-3. (Note that the polygons are still not anti-
aliased as they need further treatment.) The actual details of calculating the
intensities of different pixels can be complicated and usually results in slower
 rendering time. Refer to [FOLE95] and [WATT931 for more details on aliasing
and techniques on how to avoid it. Because anti-aliasing is a complex and
expensive operation, it is usually deployed only on an as-needed basis.

Summary
In this chapter, we have covered most of the groundwork to understand the work-
ings of the computer and how the computer stores and displays simple graphics
primitives. We have discussed the color model and how OpenGL employs the
RGB mode to set pixel colors. We have also seen how to identify pixel
coordinates and light up different points on the computer window using
OpenGL. Finally we have learned how basic shapes are rasterized, anti-aliased
and finally displayed on a grid of pixels. In the next chapter, we will explore how
to construct and move the shapes around in our 2D world.
Chapter 2
Making Them Move

In the previous chapter, we saw how to draw basic shapes using the OpenGL
graphics library. But we want to be able to do more than just draw shapes: we
want to be able to move them around to design and compose our 2D scene. We
want to be able to scale the objects to different sizes and orient them differently.
The fbnctions used for modifying objects such as translation, rotation, and
scaling are called geometric transformations.
Why do we care about transformations? Usually we define our shapes in a
coordinate system convenient to us. Using transformation equations enables us
to easily modify object locations in the world coordinate system. In addition, if
we transform our objects and play back the CG images fast enough, our eyes will
be tricked to believe we are seeing motion. This principle is used extensively in
animation, and we shall look into it in detail in Chapter 7. When we study 3D
models, we will also see how we can use transformations to define hierarchical
objects and to "transform" the camera position.
This chapter introduces the basic 2D transformations used in CG: translation,
rotation, and scaling. These transformations are essential ingredients of all
graphics applications.
In this chapter you will learn the following concepts:

Vectors and matrices

2D Transformations: translation, scaling, and rotation
How to use OpenGL to transform objects
Composition of transforms
2.7 Vectors and Matrices
Before we jump into the fairly mathematical discussion of transformations, let us
first brush up on the basics of vector and matrix math. This math will form the
basis for the transformation equations we shall see later in this chapter. We
discuss the math involved as applied to a 2D space. The principles are easily
extended to 3D by simply adding a third axes, namely, the z- axis.

Vectors
A vector is a quantity that has both direction and length. In CG, a vector
represents a directed line segment with a start point (its tail) and an end point (the
head, shown typically as an arrow pointed along the direction of the vector). The
length of the line is the length of the vector.

Fig. 2.1: A 2D VectorlPoint: A directional line

A 2D vector that has a length of x units along the x-axis and y units along the
y-axis is denoted as
[;I
It is valuable to think of a vector as a displacement from one point to another.
Consider two points P(1,-1) and Q(3,-2 , as shown in Figure 2.1. The
displacement from P to Q is the vector V =
[:I
, calculated by subtracting the

coordinates of the points individually. What this means is that to get from P to
Q, we shift right along the x-axis by two units and down the y-axis by one unit.
Interestingly, any point P1 with coordinates (x,y) corresponds to the vector VPI,
with its head at (x,y) and tail at (0,O) as shown in Figure 2.1. That is, there is a
one-to-one correspondence between a vector, with its tail at the origin, and a
by the vector [;i
point on the plane. This means that we can also represent the point P(x,y)

Often, the ma of transformation equations uses the vector representation of

points in this manner, so do not let this usage confuse you.

Operations with Vectors

Vectors support some fundamental operations: addition, subtraction, and
multiplication with a real number.
Vectors can be added by performing componentwise addition. If V1 is the
vector (xl,yl) and V2 is the vector (x2,y2) then V1+V2 is

Conceptually, adding two vectors results in a third vector which is the

addition of one displacement with another.

Fig. 2.2: Adding two vectors

Multiplying a vector by a numbers results in a vector whose length has been

scaled by s. For this reason, the number s is also referred to as a scalar. If s is
negative, then this results in a vector whose direction is flipped as well.

Fig. 2.3: Multiplying a vector with a scalar

 Mathematically, the scaled vector sV =

Subtraction follows easily as the addition of a vector that has been flipped:
that is VI-VZ= VI+(-V2).

Fig. 2.4: Subtracting two vectors

The Magnitude of a Vector

The length of a vector V= r x l
L1J'
is also referred to as its magnitude.
The magnitude of a vector is the distance from the tail to the head of the
vector. It is represented as IVI and is equal to .\j*ly
It is very useful to scale a vector so that the resultant vector has a length of
1, with the same direction as the original vector. This process is called
normalizing a vector. The resultant vector is called a unit vector. To normalize a
vector V, we simply scale it by the value l/(VI. The resultant unit vector is
represented as V=V/IVI.
Interestingly, a unit vector along the x-axis is quite simply the vector

Fig. 2.5: A Unit Vector

and a unit vector along the the y axis is

The Dot Product

The dot product of two vectors is a scalar quantity. The dot product is used to
solve a number of important geometric problems in graphics. The dot product is
written as V1 .V2 and is calculated as
You can find many of these vector functions coded in a utility include file
provided by us called utils.h.

Matrices
A matrix is an array of numbers. The number of rows (m) and columns (n) in the
array defines the cardinality of the matrix (m x n). In reality, a vector is simply a
matrix with one column (or a one-dimensional matrix). We saw how displays are
just a matrix of pixels. A frame buffer is a matrix of pixel values for each pixel
on the display.
Matrices can be added component wise, provided they have the same cardinality:
all

[a21
a12

+ 1:: b12

b2,]=
a l l + b l l a12+b12

a22b2J
Multiplication of a matrix by a scalar is simply the multiplication of its compo-
nents by this scalar:

Two matrices can be multiplied if and only if the number of columns of the first
matrix (m x n) is equal to the number of rows of the second (n xp). The result is
calculated by applying the dot product of each row of the first matrix with each

:I
column of the second. The resultant matrix has a cardinality of (m x p).
b l l b12
-
-
b21 b22

b3 1 b32

a l l * b l l + a12*b21 + a12*b31
a21*b11 + a22*b21 + a23*b31
all*b12 + a12*b22 + a13*b32
a21*b12 + a22*b22 + a23*b32 1
We can multiply a vector by a matrix if and only if it satisfies the rule above.

a l l a12 a13

a21 a22 a23

 An identity matrix is a square matrix (an equal number of rows and columns)
with all zeroes, except for 1s in its diagonal.

Multiplication of a vectorlmatrix by an identity matrix has no effect. Prove this

by multiplying a vector with the appropriate identity matrix. Refer to LENG93
for more details on vectors and matrices.

Whew! After that fairly exhaustive review of matrices and vectors, we are ready
to get into the thick of transformations.

2.2 2 0 Object Transformations

The functions used for modifying the size, location, and orientation of objects or
of the camera in a CG world are called geometric transformations.
Transformations are applied within a particular space domain, be it an object
space or the camera space or a texture space. We shall mainly be discussing
transformations applied in the object space, also called object transformations or
model transformations. The mathematics for transformations in other spaces
remains the same.

Object Space
Usually, objects are defined in a default coordinate system convenient to the
modeler. This coordinate system is called the object coordinate system or object
space.
Y

z I
Fig.2.5: Object space
 The object coordinate system can be transformed, in order to transform the
object within the world coordinate system. An object transformation sets the
state of the object space. Internally, this state is represented as a matrix. All the
objects drawn on the screen after this state is set are drawn with the new
transformations applied. Transformations make it convenient to move objects
within the world, without actually having to model the object again!

Object
Coordinate
System

Coordinate
System
Fig.2.6: Object space is transformed within the world coordinate system.

Let us look into the three kinds of transformations most commonly used
CG: translation, rotation, and scaling.

Fig.2.7:Translation along the x- and y-axes

Translation
Translation is the transform applied when we wish to move an object. In a 2D
world, we can move the object from left to right (translate along the x-axis) or
up and down (translate along the y-axis) as shown in Fig.2.7. The abbreviations
for these translations are Tx and Ty.
Consider a circular shape, P, as shown in Figure. If we wish to move P by a
distance of (tx,ty) then all points P(x,y) on this shape will move to a new location
P'(x,y) = &+&, p t y ) .

[:,I
If we define the vector for a given points P(x,y) as:

P = 1 and the translation vector as T= then the resultant

transformed point P'is the vector represented as P Y =

[; :];
Or more precisely
P' = T+P
Mathematically, we can translate an object by applying this equation to every
point on the object. Usually, we can get away with just applying the translation
to the vertices of the shape and then recalculating the translated shape.
Let us visually display this concept with an example of a bouncing ball. From
the last chapter, you may recall how to draw a circle:

Wefine PI 3.141 5926535898

11 cos and sin functions require angles in radians
-
11 recall that 2PI radians 360 degrees, a full circle

-
GLint circlegoints 100;
void MyCircle2flGLfloat centerx, GLfloat centery, GLfloat radius){
GLint i;
GLdouble angle;
glBeginlGL-POLYGON);
-
for (i 0; i < circlegoints; i+ + 1 {
-
angle Z*Pl*ilcircle~oints; I1 angle in radians
glVertex2flcenterx+radius*coslangle),
centery + radius*sin(angle));
1
glEnd0;
1
In Example2-1, we use this function to draw a brown ball centered at the
origin with the radius of the ball set to be 15 units in our world coordinate
system:
GLfloat RadiusOfBall - 15.;
11 Draw the ball, centered at the origin
void draw-ball0 {
glColor3f10.6,0.3,0.1;
MyCircle2flO.,O.,RadiusOfBalll;
1
We want to bounce this ball up and down in our display window. That is, we
wish to translate it along the y-axis. The floating-point routine for performing
translations in OpenGL is

glTranslatef(Tx, Ty, Tzl

It accepts three arguments for translation along the x-, y-, and z-axes
respectively. The z-axis is used for 3D worlds, and we can ignore it for now by
setting the Tz value to be 0.
Before we apply a new translation, we need to set the transformation state in
the object space to be an identity matrix (i.e., no transformation is being applied).
The command

clears out the current transformation state with the identity matrix. We do this
because we do not wish to reuse old transformation states.
Then, we constantly add (or subtract) to the ty component along the y-axis,
and draw the ball in this newly transformed position. The snippet of code to
translate an object in a window with the maximum extent of the world
coordinates set to (160,120) is shown below. A ydir variable is used to define
whether the ball is bouncing up or down (and hence whether we should
increment or decrement ty).

I1 160 is max X value in our world

I1 Define X position of the ball to be at center of window
-
xpos 80.;
11 120 is max Y value in our world
I1 set Y position to increment 1.5 times the direction of the bounce
-
ypos ypos+ydir "1.5;
I1 If ball touches the top, change direction of ball downwards
if (ypos -- - 120-RadiusOfBalll
ydir -1;
11 If ball touches the bottom, change direction of ball downwards
else if (ypos cRadiusOfBall1
-
ydir 1;
llreset transformation state
glLoadldentityO;
I1 apply the desired translation
glTranslatef(xpos,ypos, 0.1;

Camera (or viewing) and object transformations are combined into a single
matrix in OpenGL, called the model-view matrix (GL-MODELVIEW). The
command

specifies the space where the transformations are being applied. The mode is
normally set in the reshape function where we also set the clipping area of the
window.
That is, whenever the window is reshaped, resized, or first drawn, we specify
any further transformations to be applied in the object space.
If you try to view this animation, we shall see motion that is jerky, and you
may even see incomplete images being drawn on the screen. To make this
bouncing motion appear smooth, we make use of a concept called double
buffering.

Double Buffering
Double buffering provides two frame buffers for use in drawing. One buffer, called
the foreground buffer is used to display on the screen. The other buffer, called the
background buffer, is used to draw into. When the drawing is complete, the two
buffers are swapped so that the one that was being viewed is now being used for
drawing and vice versa. The swap is almost instantaneous. Since the image is
already drawn when we display it on screen, it makes the resulting animation look
smooth, and we don't see incomplete images. The only change in the required to
activate double buffering is to specify the display mode to be GLUT-DOUBLE.

This call is made in the main function of Exarnple2-1. By default, in double

buffer mode, OpenGL renders all drawing commands into the background
buffer. A call to

will cause the two buffers to be swapped. After they are swapped, we need to
inform OpenGL to redraw the window (using the new contents of the foreground
buffer).
The function
glutPostRedisplay0
forces a re-draw of the window.

The code required to display the bouncing ball is as follows:

void Displaylvoid)
{
glClearlGL-COLOR-BUFFER-BIT);
glClearlGL-COLOR-BUFFER-BIT);
11I 6 0 is max X value in our world
11Define X position of the ball to be at center of window
-
xpos 80.;
11120 is max Y value in our world
11set Y position to increment 1.5 times the direction of the bounce
-
ypos ypos+ydir "1.5;
11If ball touches the top, change direction of ball downwards
if lypos -- - 120-RadiusOfBall)
ydir -1;
11If ball touches the bottom, change direction of ball downwards
else if lypos 4adiusOfBall)
-
ydir 1;
llreset transformation state
glLoadldentii0;
11apply the translation
glTranslateflxpos,ypos, 0.);
I1draw the ball with the current transformation state applied
draw-ball0;
11swap the buffers
glutSwapBuffers0;
11force a redraw using the new contents
glutPostRedisplay0;
1
The entire code for this example can be found under Exarnple2-I/Example2_I.cpp

Scaling
An object can be scaled (stretched or shrunk) along the x- and y- axis by
multiplying all its points by the scale factors Sx and Sy. All points P=(x,y) on the
scaled shape will now become P' = (x',y') such that x' = Sx.x, y' = Sy.y. In matrix

In Figure 2.8, we show the circular shape being scaled by (112,2) and by (2,112).
 Notice from the figure that scaling changes the bottom (base) position of the
shape. This is because the scaling equation we defined occurs around the origin

Fig.2.8: Scaling of a circle.

of the world coordinate system. That is, only points at the origin remain
unchanged. Many times it is more desirable to scale the shape about some other
predefined point. In the case of a bouncing ball, we may prefer to scale about the
bottommost point of the ball because we want the base of the shape to remain
unchanged. In a later section, we shall see how to scale the shape about a
predefined position.
Let us go back to our bouncing ball example. When the ball hits the ground
we want it to squash on impact and then stretch back up again into its original
shape. This is what we expect from a real (soft) ball bouncing on the ground. The
floating point command to scale an object is

glScalef(Sx, Sy, Sz)

which accepts three arguments for scaling along the x-, y- and z-axes. Scaling has
no effect when Sx=Sy=Sz=l. For now, we will set the Sz to be 1.
We can modify the Display code from the previous example to include
scaling of the ball. When the ball hits the ground, and for some time afterwards,
we stop translating and squash the shape. That is, we scale down the shape along
the y-axis and proportionately scale up along the x-axis. When we reach a
predetermined squash, we stretch back up again and restart the translation.
Shown below is the code to perform this transformation.

11 Shape has hit the ground! Stop moving and start squashing down and then back up
if (ypos- -sy - syesquash ; - -
RadiusOfBall && ydir -1 1 {
 a (SY < 0.81
11 reached maximum squash, now un-squash back up
squash 1.1;
else if (sy > 1.I{
-
11reset squash parameters and bounce ball back upwards
- -
sy 1.;
squash 0.9;
-
ydir 1;
1
sx - 1.lsy;
1
11 120 is max Y value in our world
11set Y position to increment 1.5 times the direction of the bounce
else {
- --
ypos ypos+ ydir *1.5 - (1.-syl*RadiusOfBall;
if lypos 120-RadiusOfBalll
ydir -1; -
else if (ypos <RadiusOfBalll
ydir 1; -
1
glLoadldentity(1;
glTranslatef(xpos,ypos, 0.1;
glScalef(sx,sy, 1.I;
draw-ball();

The entire code can be found under Example 2_2/Exarnple 2-2.cpp. Notice
that two transformations are applied to the object-translation and scaling.
When you run the program, you will notice that the ball doesn't stay on the
ground when it squashes! It seems to jump up. This happens because the scaling
is happening about the origin-which is the center of the ball!

Rotation
A shape can be rotated about any of the three axes. A rotation about the z-axis
will actually rotate the shape in the xy-plane, which is what we desire in our 2D
world. The points are rotated through an angle about the world origin as shown
in Fig.2.9. Mathematically, a rotation about the z-axis by an angle 0 would result
in point P (x,y) transforming to P' (x',y') as defined below:

In matrix form:
or P=R.P
 Fig.2.9: Rotation of shape

Where R is the rotation matrix shown abovc. Positive angles are measured
counterclockwise from x toward y. Just like scaling, rotation occurs about the
world origin. Only points at the origin are unchanged after the rotation
transformation.

The OpenGL routine for rotation is

It expects the angle of rotation in degrees and a nonzero vector value about
which you wish the rotation to occur. For rotations in the xy-plane, we would set
vx=vy=O and vz =1 (i.e., the unit vector along the z-axis)
Let us go back to our bouncing ball example. Of course, rotating a round ball
will have no effect on the ball. Let us redefine the draw-ball routine to draw an
clongated ball.

11 Draw the ball, centered at the origin and scaled along the X axis
11 It's a football!
void draw ball0 {
gl~olor3f(0.6,0.3,0.~;
glScalef(l.3,I .,.I;I
MyCircle2f(O.,O.,RadiusOfBalI);
1
To rotatc the ball while it's bouncing, we add thc following lines of code to
our Display function:

rot - rot+2.;
 I/ reset rotation after a full circle

--
if (rot > 3601
rot 0;
glLoadldentity0;
glTranslateflxpos,ypos, 0.1;
glScalef(sx,sy, 1.I;
glRotateflrot, 0.,0.,1 .I;
draw-ball0;

The entire code can be found under Example 2_3/Example 2-3.cpp.

Note that we apply the transformations in a certain order-first rotate, then scale,
and finally translate. The order of transformations does affect the final outcome
of the display. Try changing the order in the example and see what happens.
Let us see how transformations are combined together in morc detail.

2.3 Homogenous Coordinates and Composition of

Matrix Transformations
We have seen the different vectorlmatrix representations for translation, scaling,
and rotation. Unfortunately these all differ in their representations and cannot be
combined in a consistent manner. To treat all transformations in a consistent way,
the concept of homogenous coordinates was borrowed from geometry and
applicd to CG.
With homogenous coordinates, we add a third coordinate to a (2D) point.
Instead of being represented by a pair of numbers (x,y), each point is now
represented by a triplet (x,y, I+'),or in vector notation as r x i

14
To go from homogenous coordinates back to our original non-homogenous
world, we just divide the coordinates by W. So the homogenous point
represented by:

I;[
is equal to the point (x/K y/I+').

Two points of homogenous coordinates (x,y, W) and (x ',y ', W') are the same point
if one is a multiple of the other. So, for example, (1,2,1) and (2,4,2) are the same
points represented by different coordinate triples. The points with W=O are points
at infinity and will not appear in our discussions.
Because 2D points are now three elcmcnt column vcctors, transformation
matrices used to transform a point to a new location must be of cardinality 3x3.
In this system, the translation matrix T is defined as
 Any point P (x,y,W) that is translated by the matrix T results in the
transformed point P' defined by the matrix-vector product of T and P:

Satisfy yourself that for W=l this is consistent with what we learned earlier.
For any other value of W, convert the coordinates to their non-homogenous form
by dividing by W and verify the result as well.

The Scale transformation matrix (S) in homogenous coordinates is defined

as:

S=

and the rotation matrix (R) as:

sin@) cos(8) 0 /
Each of these matrices can be multiplied successively by the homogenized
points of our object to yield the final transformed points. For example, suppose
that points on shape P are translated by a translation matrix T1 and then by T2.
The net result should be the translation (T1+T2). Verify that the equation:

does indeed result in the desired transformed points.

In fact, it can be shown that this equation can be derived to be:

That is, the matrix product of the two translations produces the desired
transformation matrix. This matrix product is referred to as the concatenation or
composition of the two transformations. Again, please verify for yourself that
this is indeed the case. It can be proven mathematically that applying successive
transformations to a vertex is equivalent to calculating the product of the
transformation matrices first and then multiplying the compounded matrix to the
vertex [FOLE95]. That is, we can selectively multiply the fbndamental R, S and
T matrices to produce desired composite transformation matrices.
 The basic purpose of composing transformations is to gain efficiency. Rather
than applying a series of transformations one after another, a single composed
transformation is applied to the points on our object. At any given time, the
current transformation state is represented by the composite matrix of the applied
transformations.
Let us apply the composition principle to the ball shape under discussion. Let
us go back to Exarnple2-2, where we were squashing the ball upon contact with
the ground. Remember that the ball seemed to jump when we squashed it. To
avoid this jump, we wish to scale the ball about its bottom most point. In other
words, we wish the bottommost point to stay at the same place when we squash
the ball.

I
I
I
I
I
I Fix
T1

l'l(xl,~l)
--
Fig.2.10: Translate shape by T I

From our study of transformations, we know how to scale the ball about the
world origin (which is where the center of the ball is located). To transform the
ball about its base, we can convert this problem into three basic transformations:
First, move the ball so that the desired fixed point of scaling (the base in this

I pl (xl,yl)
Fig.2.11: Scaling the translated shape
case) point Pl (XI,yl) is at the origin, by applying translation matrix TI. Then we
scale the shape as usual by applying a scaling transform S. The scaling occurs
about the origin, which is now where the base of the ball is positioned.
Finally, we translate the shape back such that P1 returns to its original point,
i.e. we apply a translation transform T2 = -TI. Viola: we have scaled the shape
keeping the bottommost point P1 (x1,yl) fixed. Effectively, we have scaled the
shape about the pivot point P I . Rotating a shape about a point works on the same
principle. First translate the shape so that the point of rotation is at the origin.
Apply the rotation and translate the shape back to its original location.

Fig.2.12: Translate shape back to original location.

The composite matrix to compute this entire transformation can be calculated
as
M = T2.S.T1
This composite matrix can be applied to the shape in one step to achieve the
desired result. In general, the order of composing the transforms is important. If
we change the order, different results will appear.
Let us look into how to implement matrix compositions using OpenGL. All
viewing and modeling transformations in OpenGL are represented as 16-element
arrays (representing the 4 x 4 matrix values). The element in the ith row and jth
column is represented by M[i+j*4]. For 2D transforms that use 3x3 matrices, we
simply start with a 4x4 identity matrix and then replace only those locations that
we are interested in.
The command

allows you to load a matrix directly to represent the current transformation state.
The command
multiplies the current transformation matrix by the specified matrix, M. Each
successive glMultMatrix*() or transformation command multiplies a new 4 x 4
matrix M to the current transformation matrix C to yield the (new current)
composite matrix C.M. In the end, vertices v are multiplied by the current matrix
to yield the transformed vertex locations. This process means that the last
transformation command called in your program is actually the first one applied
to the vertices. For this reason, you have to specify the matrices in the reverse
order of the transformation applied.
Consider the following code sequence, which draws a single point using
three transformations:

glMatrixModeIGL-MODELVIEW);
glLoadldentity0;
glMultMatrixflT2); 11apply transformation T2
glMultMatrixfIS1; 11 apply transformation S
glMultMatrixf(T11; I1 apply transformation T1
glBeginlGL-POINTS);
glVertex3f(v); I1 draw transformed vertex v
glEnd0;

Remember, we do need to set the matrix mode to specify which space the
matrix operations are occurring within (in this case, the object space represented
by the modelview matrix). With this code, the modelview matrix successively
contains I, T2, T2 S, and finally T2.S.T1, where I represents the identity matrix.
The final vertex tiansformation applied is equal to: (TZ.(S.(TI.v))) - notice that
the transformations to vertex v effectively occur in the opposite order in which
they were specified.
Coming back to our example, we want to scale the ball about its base. To do
this, we translate the ball so that its base lies at the origin. This is done by
translating the ball upward by its radius. If we initially define the transformation
matrix T1 as the identity matrix
-
GLfloat T1[161 {1.,0.,0.,0.,\
O.#I .ro.ro.J
o.ro.ll .,O.#\
o.ro.ro.ll .}

then to attain the matrix to translate the ball upward by its radius, we can just
set

-
T I [I31 RadiusOfBall;
Defining T1 to be { I .,0.,0.,0.,\
0.,1 .#O.#O.#\
46 2 3 HOMOGENOUS
COORDINA~ES OF MATRIX
AND C~ILIPOSITION TRANSFORMA~IONS

To move the ball back down, we set

If we define the scaling matrix S also to initially be the identity matrix, then to
squash the ball by sx and sy, S would be set as
-
S[Ol sx;
-
S[51 sy;

The transformation sequence for scaling and bouncing the ball would be as
follows:

//Retain the current position of the ball

T[121- xpos;
-
T[131 ypos;
glLoadMatrixflT);
//Squash the ball about its base

-
T I [I31 -RadiusOfBall;
I/ Translate ball back to center
glMultMatrixf(T1 I;
S[Ol sx;-
S[51 sy;-
I/ Scale the ball about its bottom
glMultMatrixf(S1;
-
T I [I31 RadiusOfBall;
I/ Translate the ball upward so that it's bottom point is at the origin
glMultMatrixflT1);
draw-ball0;

Now you will see that the ball remains on the ground when it is squashed-
just as we wanted! The entire code can be found under Example 2-4.

An Easier Alternative
The three commands we learned earlier, namely, glTranslate, glScale, and
glRotate, are all equivalent to producing an appropriate translation, rotation, or
scaling matrix and then calling glMultMatrix*() with this matrix. Using the
above functions, the same transform sequence for the bouncing ball would be as
follows:
llreset transformation state
 glLoadldenti0;
I1 retain current position
glTranslateflxpos,ypos, 0.1;

11 Translate ball back to center

glTranslateflO.,-RadiusOfBall, 0.1;
I1 Scale ball about its bottom
glScaleflsx,sy, 1.I;
I1 Translate the ball upward so that it's bottom is at the origin
glTranslateflO.,RadiusOfBall, 0.1;
I1 draw the ball
draw-ball0;

Matrix math is very important to the understanding of graphics routines. But

using matrices in actual code is involved and often tedious. It is much easier to
use the glTranslate, glScale, and glRotate commands. These commands
internally create the appropriate matrix and perform the necessary computations.
For the rest of the book, we shall use the above-mentioned OpenGL functions in
our coding examples.

Summary
In this chapter, we have covered one of the most mathematical concepts in Com-
puter Graphics: transformations. Transformations are based on vector and matrix
math. The basic transformations are of three types: translation, scale, and
rotation. These can be composed together to create complex motions. Although
the actual matrix math for transformations can be fairly tedious, OpenGL
provides a number of functions to easily manipulate and apply transformations
to objects. In the next section, we shall extend the concepts of 2D
transformations learned here to the more general 3D case. This is when we will
really be able to appreciate the true power of transformations.
Chapter 3
Pixels, Images and
Image M e s

In Chapter 1, we saw that the computer display can be treated as a two-

dimensional grid of pixels. Pixels can be addressed by their (x,y) coordinates
which represent their horizontal and vertical distance from the origin. All shapes
(even three dimensional ones) ultimately need to identify the location of the
appropriate pixels on the screen for display-a process called rasterization. In
Chapter 2, we learned how to transform these shapes to achieve motion. We used
OpenGL to render many of the concepts we learned. The two-dimensional array
of colors that we created to define our image can be saved in a file for later use.
This file is referred to as a raster image file.
Images need not be created only by the computer. Digital cameras, scanners,
etc. can all save images to a file and send the file to the computer. The real value
of saving images on the computer is what can be done to the image once it
resides on the computer. You may have played with photo editors, which let you
manipulate your photos-to reduce red eye for example. With good image
processing tools, there is no end to the magic you can do. Photoshop by Adobe
provides one of the most sophisticated image processing packages on the market.
In this chapter, we will learn the following concepts:

Image files (in particular, the BMP file format)

How to save your OpenGL images to a file
How to view image files within your OpenGL program
How to manipulate pixel data from an image
3.7 Raster Image Files
A raster image file is a file that stores the mapping of pixel coordinates to color
values. That is, the file saves the color at each pixel coordinate of the image. The
saved image files can then be used for various purposes such as printing, editing,
etc. Some commonly used image file types are: BMP, TIFF, GIF, and P E G files.
A raster file contains graphical information about the colors of the pixels. It
also contains information about the image itself, such as its width and height,
format, etc. Let us look into the details of a commonly used raster file format
called BMP.

The BMP File Format

Tlie BMP file format is the most common graphics format used on the Windows
platform. The motivation behind creating the BMP format was to store raster
image data in a format independent of the color scheme used on any particular
hardware system. The color schemes supported in BMP are monochrome, color
indexed mode, and RGB mode. Support for a transparency layer (alpha) is also
provided. Let us look into the structure of a BMP file. The data in BMP files is
stored sequentially in a binary format and is sometimes compressed. Table 3.1
shows the basic format of the bmp file which consists of either 3 or 4 parts.

e BMP file
4
- Palette (indexed mode)
Color
Pixel data
Table 3.1: Parts of the BMP file

The first part of the file is the header. The header contains information about the
type of image (BM, for BMP), its size, and the position of the actual image data with
respect to the start of the file. The header is defined in a structure as shown below:

typedef struct tagBlTMAPFlLEHEADER {

WORD bffype; /I must be "BM"
DWORD b f S i ; I/ size of file
WORD bfResenred1;
WORD bfResenred2;
DWORD bf0ffBits; //offset to start of file
) BITMAPFILEHEADER
The second part of the BMP file is the image information section. Information
such as the image width and height, type of compression and the number of
colors is contained in the information header.
The image information data is described in the structure given below. The
fields of most interest are the image width and height, the number of bits per
pixel (which should be 1, 4, 8 or 24), and the compression type. Compressions
are techniques applied to files to reduce their size. Compression techniques are
especially usehl in the case of raster image files, which tend to be huge.

typedef struct tagBITMAPINFOHEADER{

DWORD biSize;
LONG biWidth;
LONG biHeight;
WORD biplanes;
WORD biBiiCount;
DWORD bicornpression;
DWORD bisizelrnage;
LONG biXPelsPerMeter;
LONG biYPelsPerMeter;
DWORD biClrUsed;
DWORD biclrlrnportant;
) BITMAPINFOHEADER

The compression types supported by BMP are listed below:

0: no compression
I: 8 bit run length encoding
2: 4 bit run length encoding
3: RGB bitmap with mask
We will assume no compression of bitmap files in this book. Refer to
BOVIOO if you would like more information on imaging and compression
techniques.
If the image is in index color mode, then the color table information follows
the information section. We will assume only RGB mode files in this book.
Last of all is the actual pixel data. The pixel data is stored by rows, left to
right within each row. The rows are stored from bottom to top. The bits of each
pixel are packed into bytes, and each scan line is padded with zeros to be aligned
with a 32-bit boundary.
Since the bitmap image format is so simple, reading bitmap files is also very
simple. The simplest data to read is 24-bit true color images. In this case the
image data follows immediately after the information header. It consists of three
bytes per pixel in BGR (yes, it's in reverse) order. Each byte gives the intensity
for that color component-0 for no color and 255 for fully saturated. A sampling
of code required to read a 24-bit image BMP file would look as follows:
 11 Read the file header and any following bitmap information...
-
if llfp fopenlfilename, "rb"))
return (NULL);
--
NULL1

11 Read the file header

freadl&header, sizeof(BITMAPFlLEHEADER),l, fp)
-
if 1header.bffype ! 'MB'I
reversed.
I1 Check for BM

{
11 Not a bitmap file
fcloselfp);
return (NULL];
1
infosize - header.bf0ffBits
siieoflBITMAPFILE
-
HEADER);
fread(*info, 1, infosii, fp)
-
imgsize (*info)->bmiHeader.biSiielmage;
11 sometimes imagesiie is not set in files
if (imgsize -- -
01
imgsize ((*info)->bmiHeader.biWidth *
(*info)->bmiHeader.biBitCount + 71 18 *
abs(1"info)-> bmiHeader.biHeight1;
freadlpixels, 1, imgsize, fp);

Microsoft Windows provides the BMP file data structures definitions in

<wingdi.h>. On a Linux platform, you may have to define the structs yourself.
The pixels variable is a pointer to the actual bytes of color data in the image.
Conceptually, you can think of it as rows and columns of cells containing the
color values of each pixel. (Can you see the analogy between this structure and
the frame buffer?)

Fig.3.1: Array of RGB values

The code required to save a BMP file given the pixel data, is analogous.
We have provided C code to read and save bmp files. It can be found in the
files: bmp.cpp and bmp.h under the directory where you installed the example
code.. The functions have the following signatures
 extem GLubyte 'ReadBiimap(constchar "filename,BlTMAPINFO "info);
extem int SaveBiimap(constchar 'filename, BITMAPINFO 'info,GLubyte 'bits);

where BITMAPINFO is a struct containing the BITMAPINFOHEADER struct and a

color map (in the case of color index mode files).
SaveBitmap requires and ReadBitmap returns a pointer to the bytes of color
information. In bmp.h, we have commented out the BMP header struct definitions.
If you use Microsoft libraries, then you will not need to define them as they are
defined in wingdi.h.

3.2 Bitmaps and Pixmaps

Thepixels variable we saw in the last section is referred to as apixmap. A pixmap is

simply a structure that holds the color information of an image in memory. This
information can then be transferred to the frame buffer for rendering on screen or can
be saved back again (presumably after being manipulated in some way) as a file.
Historically, pixmap defines a color image whereas its counterpart, the bitmap is
monochrome (black and white only). Bitmaps have only one bit for every pixel (a value
of 1 is white and a value of 0 is black). But many people use the terms interchangeably.
Let us see how we can use OpenGL to save our work to an image file.

Saving your work to an Image file

The OpenGL function,

reads a block of pixels from the frame buffer and returns a pointer to the pixel
data-yes it returns a pixmap! The exact signature for the glReadPixels function is

void glReadPixeIs(Gl.int x,GLint y,GLsizei width,GLsiiei height,

GLenum forrnat,GLenum type, GLvoid 'piuds)

The function returns the pixel data from the frame buffer, starting with the pixel
whose lower left corner is at location (x, y), into client memory pointed to be the
variable: pixels. glReadPixels returns values for each pixel
(x + i, y + j) for 0 <= i < width and 0 <= j < height.
Pixels are returned in row order from the lowest to the highest row and from
left to right in each row. Typically, one would like to read the entire content of
the frame buffer - as defined by the extents of the viewport.
format specifies the color mode of the pixmap. GL-COLOR-INDEX for color-
indexed pixels, GL-RGB for RGB pixels, and GL-BGR-EXT for RGB-mode based
BMP files (since BMP files reverses the order of the R,G and B components).
 type specifies the bit size of each color component. GL-BYTE or
GL-UNSIGNED-BYTE is used for 8-bit values (which are used by RGB-mode
images), GL-BITMAP is used for one-bit values (monochrome images) etc.
Several other parameters control the processing of the pixel data before it is
placed into client memory. These parameters are set with three commands:
glPixelStore, glPixelTransfer, and glPixelMap.We shall not look into the details
of these calls here. Interested readers are encouraged to refer to SHRE03 for
more information. For your convenience, we have defined a function:

GLubyte *
ReadBitmapFromScreenlBITMAPINFO **info)

that reads the current frame buffer, constructs an appropriate BITMAPINFO, and
returns the constructed pixmap. This function is declared in the file: bmp.h and
defined in the file: bmp.cpp. It is useful (but not necessary) to look at the code to
understand how to setup OpenGL to read pixel values from the frame buffer.
Recall the stick figure we drew in Examplel-3. In Example3-I, we save the
stick figure to a BMP file called stick.bmp in the same directory as the
executable. The code to read and save the state of the frame buffer is as follows:

BITMAPINFO *info;
-
GLubyte "pixels ReadBitmapFromScreeni&info);
SaveBitmapi"stick.bmp", info,pixels);

These lines of code are to be called after all drawing operations are complete,
in order that the frame buffer is completely defined. The generated BMP file can
now be loaded into your Microsoft Paint or any other program for further editing.
The code to draw out the stick figure and save it to a file can be found under
Example3-I/Example3~l.cpp.You will need to compile and link it with the
bmp.cpp file that we have provided.

Loading an Image file and using it in OpenGL

Let us see how we can use OpenGL to load an image file.

OpenGL provides the library function

to take a bitmap and transfer its contents to the frame buffer for drawing to the
screen. The hnction

glDrawPixelsiGLSizei width, glSizei height, Glenum format, Glenum type, Ghroid * pixels)

draws pixmaps onto the screen. It accepts five arguments:

Where width and height specifies the dimensions of the pixmap.
format specifies the color mode of the pixmap-GL-BGR-EXT for RGB
mode BMP files.
type specifies the bit size of each color component: in our case,
GL-UNSIGNEDBYTE.
And finally comes the actual pixel data.
The location of the bottom left comer of the pixmap on the application
window is determined by the most recent call to the hnction

where x and y are the world coordinates along the x- and y-axis. This position is
affected by any transformations we may apply to the current object state.
A point to note about bitmaps and pixmaps is that they are defined upside
down! You can specify a negative height to invert the image.
In Example3-2, we read in a 64 by 64-sized pixmap. The image has a black
background and a yellow colored smiley face.

Fig.3.2: A Smiley face

We keep redrawing this pixmap, bouncing it up and down in our window.

Additionally, we also bounce our ball from Chapter 2 so as to demonstrate that
we can still make valid OpenGL drawing requests.

void Displaylvoidl
{
glClearlGL COLOR-BUFFER-BIT);
gl~aster~os2f1100,~osl;
I1 Draw the loaded pixels
glDrawPixelslinfo-> bmiHeader.biWidth,
info->bmiHeader.biHeight,
GL BGR EXT, GL-UNSIGNED-BYTE, pixels);
I1 draw the ball;t th&ame Y position
draw-ballWO.,yposl;
lldont wait, start flushing opengl calls to display buffer
glFlush0;
glutSwapBuffers0;
 11 120 is max Y value in our logical coordinate system
- -
ypos ypos + 0.1;
if (ypos > 120.1
-
ypos 0;
11 Force a redisplay
glutPostRedisplayl1;
1
Notice that we call the ball drawing code using its center as a parameter. This
approach ensures that we draw the ball with a translated center without using the
OpenGL transformation functions. The reason is that the function glRasterPos is
affected by transformations. This, in turn means that it can be hard to control the
exact location of our smiley face. We shall see techniques on how to handle this
issue in a later chapter. The main () function reads in thc dcsired bitmap.

BITMAPINFO "info; 11 Bitmap information

GLubyte "pixels; 11 Actual pixel data

11 read in image file as specified on the command line

if (argc > 11
pixels - ReadBitmap(argv[ll, &info);
1
The code for this example can be found in Example3-I/ Example3-1.cpp.
Default images are located undcr the directory Images under thc installed folder
for the sample programs.
Notice thc location of the pixmap versus the ball. Why is the ball located at
a position lower than the image? Try to change the code so that the ball and the
image bouncc randomly along thc x- and y-axcs. For a really cool projcct, makc
the program exit when the two shapes collide. In the next chapter, we will see
how to make these kinds of shapcs movc based on user input.

Loading more than one image

In most games of today, you see characters moving across the screen in front of
a static background image. In Example3-3, we read in two pixmaps. One is the
smiley face we just saw, and the other is a background image of the Alps. We
make the background image cover the entire window by scaling it up using the
function

glPixelZoom~x,yl
This function will scale up the image by the specified x and y factors.
CHAPTER3 PIXELS,IMAGES
AND IMAGE
FILES

The code to draw the background p i i a p is shown below:

if lbgpixelsl {
glRasterPos2il0,Ol;
I1 scale the image appropriately and then draw the background image
glPiielZooml0.5,0.8l;
glDrawPixelslbginfo-> bmiHeader.biWidth,
bginfo-> bmiHeader.biHeight,
GL-BGR-EXT, GL-UNSIGNED-BYTE, bgpixelsl;
1
The motion of the smiley face is the same as in Example3-2. The entire code can
be found in Example3-3/Example3-3.cpp.In the above example, we redraw the
entire background image at every call to the display function. Unfortunately, this
approach leads to significantly slow performance. For complex images, the
computational and redisplay rate of the computer may not be fast enough to
display convincing motion. In the movie world, the images are saved and
replayed at a later time, so the slow re-display rate is not an issue.
If you save each image drawn into a bitmap file and name the saved files
sequentially (like testl.bmp, test2.bmp, test3.bmp), you can string the images
together using a movie editor such as QuickTime Pro. The editor strings together
the images in a sequence to generate a movie file--usually an MPEG or an
MPEG-4 file. The movie file can be transferred directly to tape, or can be played
by a movie player such as QuickTime. The movie file is optimized for playback
and since the player is not actually calculating the images, the playback is fast
enough to convey believable motion. This technique is also employed in
streaming videos over the Internet.
For real-time games, where saving images and then replaying them is not an
option, speed can be accomplished using other tricks. One technique is to re-
draw only the pixmap of the character, leaving the background as is, using what
we call overlay planes.
Graphical overlay planes are made up of additional memory positioned
logically on top of the frame buffer (thus the name overlay). Typically this

transparent pixel

0 Overlay Buffer

Fig.3.3: The overlay plane concept.

Frame Buffer
 creates an overlay buffer that does not disturb the contents of the existing frame
buffer, as shown in Fig.3.3.
Drawing and erasing in the overlay plane does not affect the frame buffer.
Anything drawn in the overlay plane is always visible, and usually a color of 0
renders the overlay pixel transparent (that is, you can see the frame buffer at these
values). As the frame buffer is drawn, the x-y-coordinates of each pixel are checked
against the overlay buffer pixel to see if it is nontransparent. If so, then the overlay
pixel is drawn instead. Popup menus are usually implemented using overlay planes.
Overlay planes are not supported natively by most standard graphics cards,
so we do not implement them in this book. Refer to [SHRE03] for more details
on how to develop code using overlay planes.
There are other ways we can get around the speed issue when performing
intensive pixel copying and erasing. A common technique is to use logical
operations on the pixels. Logical operations form the basis for many image
processing techniques, so we will devote the next section on it.

3.3 Computer Display Systems

Logical operations are performed between two data bits (except for the NOT
operation, which is performed on one). Bits can be either 1 or 0 (sometimes
referred to as TRUE and FALSE). The most basic logical operators that can be
performed on these bits are: AND, OR, XOR, and INVERTINOT. These
operators are essential to performing digital math operations. Table 2 shows the
values (truth table) for these operations.
Since pixels are nothing more than a series of bits in a buffer, they can be
combined using bitwise logical operations (that is, we apply the logical
operations to each corresponding bit). The operations are applied between the
incoming pixel values (source) and those currently into the frame buffer

Operation Value
0 AND 0 0
1 AND 0 0
1 AND 1 1
OORO 0
1OR0 1
10R1 1
0 XOR 0 0
1 XOR 0 1
1 XOR 1 0
NOT(0) 1
NOT(1)
- --- 0
Table 3.2: Logical operations
(destination). The resultant pixel values are saved back in the frame buffer. For
true color pixels, the operations are performed on the corresponding bits of the
two pixel values, yielding the final pixel value.
Logical operations are especially useful on bit-blt type machines. These
machines allow you to perform an arbitrary logical operation on the incoming
data and the data already present, ultimately replacing the existing data with the
results of the operation, all in hardware. Since this process can be implemented
fairly cheaply and quickly in hardware, many such machines are available. All
gaming hardware, such as Nintendo and Xbox, support these operations in
hardware for fast implementations of pixellbit copying and drawing.
In Fig. 3.4, we show the OR operator applied to two monochrome (single
bit) pixel values. The bits with value 1 are shown in a white color, and those with
value 0 are shown in black.

Incoming pixels Destination pixels Resultant pixels

(frame buffer) (saved back into
the frame buffer)

Fig.3.4: The OR operation

ORing the source over the destination combines the two pixel values. Zero-
value pixels in the source are effectively transparent pixels, since the destination
retains its pixel value at these points. The same operation can be performed on
RGB pixels by doing a bitwise OR operation.
An AND operator uses the source as a mask to select pixels in the destination
and clears out the rest, as shown in Figure .

Fig.3.5: The AND operation

Image compositing, where two images are combined in some manner to

create a third resultant image, make use of these two operations heavily.
XORing the source over the destination guarantees that the source stands out.
This technique is used very often in rubber-banding, where a line or a rectangle
is dragged around by the user on top of a background image.
Fig.3.6: Rubber-banding: XOR of a line with the background image
The neat thing about an XOR operation is that two XORs generates the same
values back again. That is,
(A XOR B) XOR A = B.
This operation is often used to to erase a previously drawn image. If you have
ever tried selecting a number of programs on the Windows desktop, now you

Fig.3.7: XORing twice

know that it's done are using the XOR operator.
In Example3-I and Example3-2, we had to redraw the entire background.
Instead, we can use the XOR operation to first draw the smiley face and then
XOR once again to erase only the portion we drew into. This technique tends to
be a much faster operation than redrawing the entire background and is therefore
used extensively in gaming.
A note of caution!. This approach works best for constant color images, since
the colors do get flipped on the first XOR. For multicolor images, the XOR
operation may result in psychedelic colors being produced. As a result, XOR
should be used only when the image colors are designed to work for it. Let us
see how we can modify Example3-1 to use logical operations.
 The OpenGL command to set a logical operation to be applied to incoming
values is

The operation can be of type GL-XOR, GL-AND, GL-OR or GL-NOT.

Logical operations need to be enabled before they will have effect. To do this,
call the hnction

glDisable() will disable logical operations from occuring.

In the display code shown below, we clear out the background only once at
startup. Then we use two XOR operations, one to draw the smiley face and then
another to erase it. When you run the program you will notice that this process
does flip the color of the smiley face!

if (FIRSTDISPLAY) {
11 Only clear the background the first time
glClear(GL-COLOR-BUFFER-BIT);
-
FIRSTDISPLAY FALSE;
11 enable logical operations
glEnable(GL-COLOR-LOGIC-OP);
11 Set logical operation to XOR
glLogicOplGL-XOR);
) else {
11 XOR incoming values with pixels in frame buffer
11 the next two lines will erase the previous image drawn
glRasterPos2f(prevxIprevyl;
glDrawPixels(info->bmiHeader.biWidth,
info-> bmiHeader.biHeight,
GL-BGR-UCT, GL-UNSIGNED-BYTE, piielsl;
1
I1 the next two lines draws in a new (XORed) image
glRasterPos2f~xpos,ypos~;
glDrawPixelslinfo-> bmiHeader.biWidth,
info-> bmiHeader.biHeight,
GL-BGR-UCT, GL-UNSIGNED-BYTE, pixels);

An additional bonus: since the smiley face has a black background (recall
that black is the transparent color by default), the resultant display only draws
the face, not the entire rectangular image. The entire code can be found in
Example3-4/Example 3-4.cpp. Try using the background image of the Alps and
see whether the XOR operation results in a desirable output.
3.4 lmage Enhancements
The real value of saving images on the computer is image processing - what can
be done to the image once it is resident on the computer. Almost all production
studios use a processing tool to perform final image touchups, called post-
production, to enhance the quality of their images. These images can be from livc
action footage or from computer-generated imagery. Effects like blending,
compositing, and cleaning up pixel values are used routinely in productions. The
idea behind image processing is simple: manipulate the pixel information of the
given image using some mathematical functions and save the result in a new
image file. We discuss two techniques in this section, namely, compositing and
red-eye removal.
Refer to [PORT841 for information on other image processing techniques.

lmage Compositing
Compositing is the most widely used operation in the post-production of films,
commercials, and even TV broadcasts. The basic operation in image compositing
is to overlay one (usually nonrectangular) image on top of the other. Recall from
Exarnple3-2, when we drew the smiley face on top of the background, we
displayed the entire image, not just the face, as we would have liked. XORing
flips the colors, so that is not always a good solution either. Compositing to the
rescue!
Various techniques are used for compositing images. One popular technique
is the blue screen process. First, the subject is photographed in front of an evenly
lit, bright, pure blue background. Then the compositing process, whether
photographic or electronic, replaces all the blue in the picture with the
background image, known as the background plate. In reality, any color can be
used for the background, but blue has been favored since it is the complementary
color to flesh tone. The source and destination images can come from livc action
images or be computer generated. Jurassic Park employed compositing
extensively to composite CG-generated dinosaurs in front of real livc shots.
The essential compositing algorithm is as shown below. It designates Blue or
the desired color as transparent. The source image is copied on top of a defined
destination image as follows:

-
for y 1 to height
-
for x 1 to width
if image[x, y] o transparent pixel then
copy irnageh yl
else
leave the destination unchanged
 Fig. 3.8: Cornpositing Algorithm
The only problem, of course, is that "transparent" color can't be used in the
source image. Another technique for compositing, being used more frequently
now makes use of a mask or stencil. This technique uses logical operations as
follows:
Create a mask of the image. Many production houses create the mask
as part of the alpha channel of the image. The mask essentially
identifies the desired areas of the source image.
AND the source with the mask (not always required).
AND the background with NOT of mask
OR the results to produce the final composite.

AND AND

Fig. 3.9: Image Composition using masks

We show below, sample code for compositing the smiley face on top of the
background image of the Alps. The mask was created as a BMP image using
Adobe Photoshop. The entire code can be found under Example3_5/Example
3-5.cpp.
 /I First copy background image into frame buffer
glLogicOplGL COPY);
gl~aster~os2i(0,0);
glPielZoom(0.5,0.8);
glDrawPixelslbginfo->bmiHeader.biWidth, bginfo->bmiHeader.biHeight,
GL-BGR-EXT, GL-UNSIGNED-BYTE, bgpixels);

I/ perform an AND with destination and NOTlsource - mask)

glLogicOp(GL-AND-INVERTED);
glRasterPos2flxpos,ypos~;
glPixelZooml1.0,1.0~;
glDrawPixels(maskinfo->bmiHeader.biWidth, maskinfo-> bmiHeader.biHeight,
GL-BGR-EXT, GL-UNSIGNED-BYTE, maskpixels);

I/ Perform an OR with source- smiley

glLogicOplGL-OR);
glDrawPixels(info->bmHeader.biWidth info->bmiHeader.biHeight,
GL-BGR-EXT, GL-UNSIGNED-BYTE, pixels);

You will see a composite image of the smiley face on top of the Alps-just
what we wanted!

Red-Eye Treatment
Image enhancements are possible because we have access to the actual pixel
information of the image.-we can target specific pixels based on some criteria
and then change their values. You may have used a very common feature in
image processing tools-red-eye removal.
Let us see how we can implement this feature for ourselves.
In Example3-6, we load in a smiley face with red eyes. The eyes are made
up of pixels with a color value of (255,0,0). We will target pixels with this color
and change them to blue.
To do this, we first determine the length of each row. Using the length, we
can point to any desired row in the pixmap as the pixel-row variable. Then we
loop through the columns of pixel values in this row.
If we determine the pixel value in any column has a red color, we change it
to blue. The code to do this is shown below:

I/ length of each row -

//remember that the color component values are padded to a 32 bit value
-
int length (info->bmiHeader.biWidth 3 + 3) & 3;
I/ for each row
-
-
for (row 0; row < info-> bmiHeader.biHeight; row + + 1 {
-
pixel-row pixels + row*length;
//Now we can loop through all the pixel values in this row as shown
 -
for (col- 0; c o k info-> bmiHeader.biWidth; col+ +, pixel-row + 3) {
11 Is the pixel value at this row and col red?
11 Remember!! BMP files save color info in reverse order (B,G,RI
--
if (pixel row[O] 0 && pixel rowtll
11 yes, cThange the piiel color to-blue.
-- 0 && pixel-row[21 - - 255l{
-
pixel-row[OI 255; pixel-rowt21 0;-
1

The entire code can be found in Example3-6/Example3-6.cpp. When you

run this code, the display will show a blue-eyed smiley face! You can save the
result to a file to verify that we did indeed perform red-eye removal. This is an
oversimplification of the actual mechanics behind red eye reduction, but you get
the idea. Now when you use a processing tool to get rid of the reds, you will
know what kind of technology is working in the background.

Summary
In this chapter, we have seen how images are stored on the computer. Any
kind of graphics, whether a visually realistic simulation of the environment or a
simple 2D drawing, is eventually saved as a 2D raster image on the computer.
Storing and viewing these images is just the tip of the iceberg: we can use these
images in our own projects and even modify them to achieve desired results and
special effects. The techniques we have discussed here are used extensively in
production suites for image processing and gaming. They form the basis for the
more complicated image enhancements and effects used in the industry. In the
next chapter, we shall put together all our current working knowledge to design
and develop a 2D game.
Let The Games Begin

In the last few chapters, we learned the basics of 2D computer graphics. We saw
how to draw simple shapes using OpenGL and how to assign colors to them. We
learned how to make these shapes move about in our CG world. We also saw
how to save, load, and modify computer images. Let us make things interesting
by using all our knowledge to design a 2D game.
When it comes to games, we need real-time graphics-that is, the program has
to be interactive and have an immediate response to input. You press a button,
and the graphics changes. You drag the mouse, and a character moves in
response. In this chapter, we shall learn some techniques to interact with user
input in real time.
We begin by first exploring what a computer game really is and discussing
the traditional processes used in designing one. Our discussions on game design
will pave the way for a real game implementation.
In this chapter, you will learn the following:

w What a game is
w What steps production houses follow when producing a game
w How to follow this process to design and implement your own game
 --- --- - -

4.7 What is a Game?

If we desire to design a game, we must define what is meant by the word game.
A game, like a story, can be thought of as a representation of a fantasy that the
readerlplayer is invited to experience. A story is a vehicle for representing
fantasy through the cause-and-effect relationships suggested by the sequence of
facts it details. Games attempt to represent fantasy by a branching tree of
sequences. It allows the player to create his or her own story by making choices
at each branch point.

A Story A Game

Fig.4.1: A Story vs. a Game

Game-playing requires a player. The game designer works to produce a game

so the player is enticed to play it. Hence, the human player is the primary con-
cern of the game designer. Why do people play games? What motivates them?
What makes games fun? The answers to these questions are crucial to good game
design. Although games are played using different media-board games, card
games, etc.-we are currently interested only in games played on the computer.
There are many motivations for humans to play games: learning, fantasy1
exploration, social acceptance, need for acknowledgment, etc. Some factors
motivate a person to play games; other factors help that person select a particular
game. A game cannot be fun if its factors do not satisfy the motivations of the
player. So when designing a game, it is important to understand the motivations
of the target audience. Different games motivate different kinds of audience. An
intellectual audience may be motivated to play a game for mental exercise, so
they would probably prefer those games that offer a greater mental challenge
than do other games. If the target audience is motivated by action and combat,
then that is what the designer should integrate into the game.
Sensory gratification is usually the deciding factor between games with the
 same intent. Good graphics, color, animation, and sound are all valued by game
players. These elements support the fantasy of the game by providing sensory
"proof' of the game's reality.
Keeping the above in mind, let us work through designing and implementing
a game for ourselves.

4.2 Game design

Let us work through the design of a game following the standard steps used in
professional game design. These steps force the designer to think through the
important issues about the game before beginning implementation of the game.
This is a very important process, since if the goals of the game are not clearly
identified, the game will probably fail to meet its primary objective of motivating
players.

1. Audience
The first step in our game design process is to identify the target audience. For
the purposes of this exercise, let us pick children as our audience. In particular-
boys between the ages of 8 and 12.
Once the audience is chosen, research is carried out to identify the likes and
dislikes of this group. Say we find out that among the many things the boys love
to do, aiming and shooting at objects rank high.
Let us define our game to be a shooting game. The goal of the game will be
to provide the player with the opportunity to aim and fire at objects. The game
will also enhance the player's fine motor skills and hand-eye coordination.

2. Environment
Next, we must determine the environment in which the game will be played. We
have already defined the game to be a shooting game, but what characters are
playing in the game and what environment are they set in?
Let us define the game to be set in outer space. We need a character that will be
controlled by the player. This character will allow the player to shoot. We also
need a character to shoot down. We further want these characters to blend into
the environment selected.
With this in mind, we define a user-controlled spaceship as the shooter, and
meteors that appear randomly as the target. In order for the game to have a good
story, we probably need a third character which is the object that the user is
trying to protect. A silhouette of planet Earth would be a good choice, since
everyone is motivated to protect it from being hit by meteors!

3. Interactive screenplay
The screenplay, if appropriate, contains the dialogs and the storyline imple-
mented in the game.
For this game, let us define a simple screenplay. The game opens with a space
setting simulated with stars and sky. The tip of planet Earth is at the bottom of
the scene. A spaceship is hovering on top of the-earth. The player can move the
ship left and right and shoot bullets upwards. Meteors, generated by the game,
randomly appear in the sky. The player needs to shoot down these falling
meteors. If a bullet hits a meteor, it is destroyed. If the meteor hits Earth, then
the game ends with an explosion.
The interactive logic of this game can be shown as follows:

Meteors appear

\ J
Meteor shot down?

Meteor hits
and bullet

4. Storyboard
Storyboards aid in the design process by illustrating the game in motion. Usually,
storyboards are quick sketches of key moments during the lifetime of the game.
Any quirks or inconsistencies in the game show up at this step.
Storyboard illustrate only the main points of the story. For this game, the
following key points would be illustrated (the actual illustrations are left as an
exercise to the reader):
The opening scene: The spaceship hovering:
a Meteors appearing randomly in space
User positioning spaceship and firing at meteor
a End of game when meteor hits earth

5. Implementation
Finally, implementation issues detail the choices made to develop the game.
Some of the choices are as follows:
Is the game 2D or 3D?
 We have no choice for the moment: it will be a 2D game.
What kind of look and feel does the game have?
The game is set up to be in space. We are going for a cartoonish look as opposed
to abstract or surrealistic.
How are the characters implemented?
The sky and stars can be implementcd using a background image. We can
probably get away with defining the earth to be part of this background image,
as long as we know its coordinate data in order to determine collisions.

Fig.4.2: The Background

The spaceship will also be rendered using an image file as shown.

Fig.4.3: The Spaceship

The meteors are merely point shapes that randomly appear in the sky and fall
down towards Earth. Bullets are also point shapes that shoot upwards when the
user clicks the right mouse button.
What are the input rnethod~~for the user?
In arcade games, a joystick enables the user to control the game. A joystick
works very similarly to a mouse. It allows the user to move characters left, right,
up, and down, and allows the user to press a button to fire. For our game, the
mouse will provide thc input parameters. The user drags the mousc with the lcft
button down in order to move the spaceship. Clicking on the right mouse button
controls firing the bullets.
What are the tools needed to develop the game?
C++ and OpenGL We will develop the game in C++ since this language lends
itself very nicely to game development.
What platform will the game run on?
Windows (but you can compile and run it on any platform)
4.3 Implementing the Game
By now you are probably more than ready to roll up your sleeves and start
implementing the game. Let us begin. The algorithm for the game (based on the
interactive logic chart we drew up earlier) is as follows:

till game ends do {

monitor user input to update spaceship location
monitor user input to generate bullet
randomly create meteors
for all meteors {
update meteor location
if meteor h i s earth {
Display explosion
End Game;

if meteor has gone out of screen

destroy meteor
else
draw meteor
for all bullets {
update bullet location
if bullet hits meteor{
destroy meteor
destroy bullet
) else if bullet is out of screen
destroy bullet
else
draw bullet

Let us begin developing the game by first defining the characters that play in our
game.

The Characters
Object oriented programming lends itself very nicely to game programming,
since all the functionality of the characters can be encapsulated within a class.
We shall use simple C++ classes to create the characters in the game. If you have
never used C++ before, do not despair! The code is very similar to C and you
will find it very easy to follow the logic of the code. Just think of a C++ class as
a C struct.

The Meteor object

First, let us consider the meteor object. A meteor has a location, which identifies
it's (x,y) coordinates on the screen. Based on the location, the meteor object
needs to be able to draw itself It needs to be able to update its location within
every loop of the code-sometimes referred to as a tick of the game. It also needs
to be able to identify whether it is out of the screen (in which case we destroy it)
or whether it has hit planet Earth.
The meteor class can defined as follows

class Meteor
{
public:
Meteorlint ulD);
-
virtual MeteorO;
void DrawO; //Draw the meteor
void DrawCollisionO; //Draw the meteor that has collided
void UpdateO; //Update the meteor location, for this example simply
decrement it's y-coordinate in order t o move it down
boo1 HitEarthO; //returns true if the meteor has hit Earth
boo1 OutOfScreenO; //returns true if the meteor is out of screen
GLint ID; I/ unique ID t o identify the meteor
GLfloat* GetLocationO {return location;)
private:
GLfloat location[2]; I/ meteor's location. location[Ol=x-, location[ll = y- coordinate
1;
The location array stores the current (x,y)-coordinates of the meteor. For this
example, we let all meteors fall straight down along the y-axis.
The update hnction merely decrements the location[l] variable in order to
make the meteor fall down. You can experiment with giving meteors an x-
direction as well.
The Draw routine draws a white colored point @Vertex) at the meteor's
current location, whereas the DrawCollision routine draws a bigger red colored
point to indicate a collision.
The functions HitEarth and OutOfScreen test the meteor's current location to
determine if it has hit Earth or is out of the boundary of the defined world
coorindates. It returns a TRUE value if the test is positive.
 We use the ID variable to uniquely identify each meteor.
The entirc code for this object can be found under the directory Example4-I,
in files Meteo~ h and Meteo~cpp.
The function createMeteor is used to randomly create meteors during the
game.The function generates a random numbcr betwcen 0 and 1 and only crcatcs
a meteor if this number is less than a constant. This is to ensure we do not have
too many mctcors bcing generated.

void createMeteor0 {
if I IIfloatlrandOlRAND~MM)< 0.991
return;
SYSTEMTIME systirne;
GetSystemTime(&systime);
-
int ID systime.wMinute*60+systirne.wSecond;
-
Meteors[lDl new MeteorllD); llcreate a meteor with identifier
1
Meteors are created only if a random number generated between 0 and 1 is
less than a constant value (0.99). Decreasing this constant will causc more
meteors to be created. Each meteor is instantiated with a unique ID that is an
integer value. Thc ID of the metcor is derived from the minute and second of thc
current time and is unique just as long as the player does not play the game for
more than an hour. This ID is used as a key to inscrt and locate the meteor in a
meteormap:

typedef map<int, Meteor*> MeteorMap;

static MeteorMap Meteors;

We use an STL containcr-a map in this case-for storing the meteors. STL
maps are an easy way to store objects in an array indexed by a key. In this case,
the ID is the kcy used to store and retrieve the corrcsponding mctcor objcct from
the MeteorMap. STL iterators are used to loop through STL containers. More
information on how to use STL can be found in GLAS95. You can easily changc
the code to manage your meteor container as a linked list.

The Bullet Object

The class encapsulating the bullet object is defined similarly to the Meteor class
we just saw. The Bullet class has an additional function to test for collisions
between thc bullet and a givcn meteor.

class Bullet
{
public:
Bulletlint 1.110, int x, int yl; IIDefine a new bullet at location(x,y).
 -
virtual BulletO;
void Draw(); //Draw Bullet
void UpdateO; //Update Bullet location
boo1 OutOfScreenO; //returns true if Bullet is out of screen
boo1 CollidelMeteor "I; //returns true if Bullet collides with Meteor
GLfloat ID;
private:

GLfloat location[ZI; /llx,y) location of bullet

1;
The Draw routine draws a red-colored point (glvertex) at the bullet's current
location, whereas the Drawcollision routine draws a bigger red colored point to
indicate a collision. The update function merely increments the location[l]
variable, thereby moving the bullet up.
All bullets created are entered into a BulletMap and located by their ID. The
code for this class can be found under the folder Example4-1 in files Bullet.cpp
and Bullet.h.

typedef map<int, Bullet*> BulletMap;

static BulletMap Bullets;

Bullets are created when the user clicks the right mouse button. We shall see
shortly, how to monitor for mouse events. The createBullet function is called
when the right mouse button is clicked. The function is defined as follows:

I1 Create a bullet and add it to a Bulletmap

void createBulletlint x, int yl {
SYSTEMTIME systime;
GetSystemTimel&systime~;
-
int ID systime.wMinute"60 + systime.wSecond;
Bullets[lDl- new BulletllD,x,yl;
1
The Spaceship
Last of all let us consider the spaceship object. Like the meteor and bullet, the
spaceship has a location variable to determine its (x,y)-coordinates.

class Spaceship
{
public:
SpaceShipO;
-
virtual SpaceShipO;
void SetPielslBlTMAPlNFO "ssinfo, GLubyte "sspixels); /I set the pixels and bitmapinfo
 for spaceship
void DrawO; //Draw the spaceship
void UpdatelGLfloat x, GLfloat yl; llupdate location of spaceship
GLfleat GetXLocationO{ return location[Ol;)
GLfloat GetYLocationO {return location[ll;)
void StartMoveO; llstart moving the spaceship
void StopMoveO; llstop moving the spaceship update of
spaceship will not do anything
private:

GLfloat location[21;

GLubyte "mypbtels; llpixel info of spaceship

BITMAPINFO "myimginfo; llbitmap info of spacehsip image

boo1 b-updates; llare w e updating location of spaceship

The function Setpixels is used to initialize the pixmap of the spaceship. The
Draw routine of the spaceship object draws a pixmap at the object's current
location.
The spaceship is in motion when the user clicks down the right mouse and
drags the mouse around. The moment the user releases the button, the spaceship
stops moving. The Boolean variabe b-updates keeps track of whether the
spaceship is in motion (and hence whether it's location needs be updated).
The Update function is called only when b-updates has a value of TRUE. It
takes as parameter the current Mouse location, and draws the spaceship at this
location. For this exercise, we move the spaceship only left and right, so only the
x-coordinate is required for motion.

User Input
OpenGL and the GLUT library provide a number of functions to create event-
handling threads. An event-handling thread is a separate thread of execution that
listens for events such as user input, timers etc. If an event is detected, a call is
made to the appropriate "callback" function. The callback function can then
handle the event as desired. We saw event handlers for the Display and Reshape
events in the previous chapters. Event-handling threads are the crux of all gaming
programs. The GLUT function used to create a mouse event-handling thread is

glutMouseFunc~voidl*funcl~int
button, int state, int x, int yll

glutMouseFunc sets the mouse callback for the current window to be the
function func. When a user presses and releases mouse buttons in the window,
each press and release generates a call to this function. The callback function
receives information about which button was clicked
(GLUT-RIGHT-BUTTON or GLUT-LEFT-BUTTON), the state of the button
(GLUT-UP or GLUT-DOWN), and the x,y location of the mouse click within
the window (in the world coordinate system). In our example, we define our
mouse handling callback function to be

void getMouse(int button, int state, int x, int yl

When the right mouse button is clicked up, we want to fire a bullet. The
bullet starts off at the center of the spaceship. If the spaceship was being dragged
around, we also want to stop moving it. If the button is clicked down, we want
to start moving the spaceship.

void getMouseIint button, int state, int x, int yl{

I1 If right mouse is clicked up, create a bullet located at the center of the spaceship image. Stop
spaceship from moving
--
if (button GLUT-RIGHT-BUTTON && state -- GLUT-UP) {
createBullet(spaceship->GetXLocationO+25,spaceship->GetYLocationO +25);
spaceship->StopMoveO;)
else {
11 Start spaceship motion
spaceship->StartMoveO;
1
1
The command

glutMotionFunc (void (*funcll int x, int yll

is used to set the motion callback function. The motion callback for a window is
called when the mouse moves within the window while one or more mouse
buttons are pressed. We want the location of the spaceship to be updated when the
user drags the mouse around. We define our motion callback function as follows

void getMouseMotionlint x, int yl{

I1 update spaceship when mouse is moved.
I1 spaceship motion is enabled only when button is clicked down
spaceship->Updatelx, SCREENMAXI-y I;

Notice something odd? We don't use the same y-coordinate that we receive
from the mouse click! Our world coordinates set the origin to be at the bottom
leftmost corner of the window. Y is increasing as we go up the screen. However,
Windows uses (0,O) at the top left comer of the window soy- is increasing going
down! The location of mouse events is in the coordinate system of the window
and needs to be reversed to map correctly into our world! The code for this class
can be found under Example4-1, in files SpaceShip.cpp and SpaceShip.h.

Timer Callbacks
Similar to mouse event callbacks, GLUT has the ability to register timer
callbacks. Timer callbacks get triggered at regular intervals of time-the "ticks"
of the game. This mechanism is extremely useful in games, when you want the
game logic to be called at regular ticks.

The GLUT function

void glutTimerFunc I unsigned int msecs ,void (*func)(int value), value);

registers a timer callback to be triggered in a specified number of milliseconds.

It expects a pointer to the event-handling function as a parameter, as well as any
integer value to be passed to this function. In the init() code of our program, we
register a timer callback as

glutTimerFuncl100, timer, 01;

This means that the function timer (listed below) will be called after 100
millseconds. The timer function creates meteors randomly, forces the Display
function to be called and finally makes another call to re-register the timer
callback, starting off the process all over again.

void timerlint value)

{
createMeteor0;
11 Force a redisplay, Display function contains main game logic
glutPostRedisplayl);
11 Restart the timer
glutTimerFunc(l0, timer, 0);
1
Putting It All Together
Now that we have all the pieces defined, let us look into the guts of the main
code. The main ~ i s ~ l a y (function
)- follows the logic of the ;lgorithm listed
previously. The exact C++ implementation can be found in
Example4-1/Example4-1.cpp. If a meteor m hits Earth, we call the function

void EndProgramlMeteor "ml

 This function first displays a colliding meteor (a dot displayed in a red color),
then loads in a series of images to depict the Earth exploding, and finally exits
from the program. The initialization calls that set up all our event handlers, seed
the random number generator, and create the spaceship are defined in the
function init:

void initlvoidl{

/I Timer callback
glutTimerFunc(100, timer, 0);
/I Define mouse callback function
glutMouseFunclgetMousel;
/I Define Mouse Motion callback function
glutMotionFunclgetMouseMotionl;

/I random number generator seeded by current time

SYSTEMTIME systime;
GetSystemTime(&systime);
/I Seed random number generator
srand(systime.wMinuteW + systime.wSecondl;

//Create the spaceship

createSpaceShip0;

All the classes and functions can be found under the directory, Example4-1. All
the images can be found under the directory, ImageslChapter4.
The program is defined in the simplest manner possible for easy readability.
It is not the most efficient code. There are a number of tricks you can do to speed
up the program: Redrawing using XOR (remember, care must be taken to make
sure color patterns work during XOR), minimizing the number of loops, etc.
You can also add many more bells and whistles to get cool effects, such as an
animation of the planet exploding instead of a display of a few static images, a
fancier meteor object, etc. We leave this part of the game as an exercise for the
reader.

Summary
In this chapter, we have combined all of our working knowledge in Computer
Graphics to develop a computer game. We have learned the fundamentals of
game design and the processes involved in designing and developing a game. In
chapter 11, we will have the opportunity to design a 3D game.
Section 2
It's 3 0 Time!
We hope you have enjoyed your ride through the 2D world. With a solid
understanding of the 2D math involved in this world, we are ready to venture
into the world of 3D graphics.
The mathematics in a 3D world is an elegant extension of its 2D counterparts.
Most concepts in 3D are developed from 2D by simply adding a third (z-) axis.
Positions of objects are now described as (x,y,z) triplets instead ofjust (x,y) pairs.
Transformation equations are defined against three axes, not two.
But 3D graphics is a lot more than just drawing simple objects. This section
begins our quest into visual realism: how can we simulate a 3D world in the
computer and then compute images that render out like photographs?
3D Productions are built up on three basic building blocks: modeling,
rendering and animation. The first two blocks-modeling and rendering-form
the basis for this section. Together they determine the final displayed image.
In Chapters 5 and 6, we will study the basics of modeling and rendering, and
in Chapter 7, we will introduce you to some of the more advanced concepts in
the field. Chapter 8 is entirely devoted to Maya, the most popular 3D tool used
in the graphics industry.
By the end of the section, you will be able to define a 3D world and render
it on the computer to produce stunning imagery. So hold your breath and get
ready for the 3D roller coaster.
Chapter 5 -- -

3 0 Modeling

The first and most critical pillar of 3D graphics is modeling. Modeling is the
process of creating a 3D model in the computer.
In lay terms, a model is a representation of a concrete or abstract entity. This
representation can be of various kinds. Quantitative models use equations to
represent and describe system behavior; organizational models use hierarchies to
represent classification schemes. A CG model refers to the geometrical
representation of the entity. The purpose of the model is to allow people to
visualize the structure of the entity being modeled.
When we model an object on the computer, we tell the computer about the
shape, spatial layout, and connectivity of components that compose the object.
In the last section, we saw examples of how to create simple 2D models such as
polygons and circles. In this chapter, we expand our horizons to a 3D world. We
will learn the process of viewing three-dimensional worlds and how to create,
compose, and transform models in this world.
In this chapter, you will learn the following concepts
w Representation of the 3D system
3D math: vectors, matrices and transformations
Creating models of
primitive shapes
generic shapes
w Viewing the 3D world
w Creating hierarchical models using transformations

We begin our discussions by first extending our knowledge of the 2D system to 3D.
5.7 The 3 0 System
We can relate to a three-dimensional space because we see our own world in 3D.
Not only do objects in our world have length and height, as they do in the 2D
space, but they also have depth associated with them. They can also be located
closer or farther away from us.
In order to represent a 3D world, we need to extend our coordinate system into
three dimensions. Every point in the 3D world is located using three coordinate
values instead of 2. In order to define points with three coordinates, we define a third
axis, typically called the z-axis. The z-axis is perpendicular to the x- and y- axis (i.e.,
it is at 90 degrees with respect to the two axes - also referred to as orthogonal). All
three axes meet at the origin defmed as (0,0,0) as shown in Fig.5.1. This is our 3D
world coordinate system. In this book, we follow a right-handed coordinate system,
which means that the positive z-axis points out of thescreen and towards us.

Fig.5.1: The three dimensional coordinate system

All points and vector representations that we learned about in chapter 2 can be
extended to 3d by adding a third component, z. Points in 3D space are identified by
a triplet of (x,y,z) values. The vector notation for this triplet is represented as:

In Chapter 1, we saw how objects are defined by the key (2D) vertices
El
defining its shape. 3D vertices define the shape of on object in 3 Dimensions.
The OpenGl function to define a vertex point (x,y,z) in 3D is

Using this function, we can modify our circle-drawing function from Chapter
1, to the 3D world. In the following code, we define the points of the circle in
the x-z plane (not the x-y plane).
 Fig. 5.2: Circle bn the XZ plane
void MyCircle3flGLfloat centerx, GLfloat centery, GLfloat centen, GLfloat radius){
GLint i;
GLdouble theta;
glBeginlGL-POINTS);
--
for li 0;i < circlegoints; i+ + l {
theta 2*PI*ilcirclegoints;
glVertex3flcenterx+ radius*cosfthetal, centery, centen+ radius*sinltheta));
1
glEndll;
1
The circle is centered at the point: (centerx, centery, centerz). The function
@Begin function is used to define the primitive being drawn. In this case, we
merely draw points along the circumference of our circle. How would you
connect the points to draw a connected circle?

3 0 Math: Vectors
A vector in 3D has the same semantics as in 2D - it can be thought of as a
displacement from one point to another. Consider two points P(1,- 1,l) and Q(3,-
2,2) as shown in Fig.5.3. The displacement from P to Q is the vector
v= (2,-1,I) = 2
[-:I
calculated by subtracting the coordinates of the points individually. This means
that to get from point P to Q, we shift right along the x-axis by 2 units, down the
y-axis by one unit and outward by one unit along the z axis.
Any point PI with coordinates (x,y,z) corresponds to the vector Vp 1 , with its
head at (x,y,z) and tail at (0,O) as shown in. That is, a 3D point is essentially a
vector with its tail at the origin (0,0,0). A unit vector is a vector that has a
magnitude of one unit. For example, the unit vector along the z-axis is:
r:iL !A
In Example 5.1, we draw unit vectors along the positive x-,y-,z-axes in red
 Fig. 5.3: vectors in 3D Space
green and blue colors respectively. The code to draw the axes is as follows:

IX Axis
glBegin(GL-LINES);
glColor3f ~1.0,0.0,0.0);
glVertex3RO,O.,O.);
glVertex3f(l,O,.O);
glEnd0;
IN axis
glBeginlGL-LINES);
glColor3f 10.0, 1.or 0.0);
gNertex3f10.,0.,0.1;
gNertex3f(O.,I ,Ol;
glEndl1;
lR axis
glBegin(GL-LINES);
glColor3f (0.0,0.0,1.0);
glVertex3f(O.,O.,O.);
glVertex3f(O.,O.,l .I;
glEnd0;

We then call the circle-drawing routine to draw two circles with unit radii,
one centered (0,0.5,0) and the other at (0,0,0):

The circle-drawing routine has additional code to vary the colors of the point
based on its location in space. The output displays the three unit axes and the two
circles. The entire code can be found in Exarnple5-I/Example5_1.cpp.This example
 z
Fig. 5.4: Two circles in 3D Shape
contains lines of code that you do not understand as of yet - specifically with
regards to setting up the 3D world. Do not worry; by the end of this chapter, it
will all start to make sense.

Operations with Vectors

The vector math that we learned in Chapter 2 can be easily extended to 3D space.
The rules for vector addition, subtraction, and multiplication rules remain the
same- since all of these operations are component wise operations, there is just
one more component added to the mix.
For example, if VIis the vector (xl,yl,zl) and Vz is the vector (x2,y2,z2), then
VI + Vz is defined as:

21 +z2
As shown in Fig.5.5, the semantics of the addition remains the same as in 2D.
Subtraction is similarly extended.

Fig.5.5: Vector addition in 3D space

Multiplying a 3D vector V in space by a scalar s essentially scales the

vector's magnitude, but does not change its direction. The process of scaling a
vector such that its magnitude is of unit length is known as normalization. There
is one vector operation not found in 2D but crucial to 3D math - that of the
vector cross product. A cross product of two vectors is best understood in terms
of planes and normal vectors.
A plane is a flat surface, much like a piece of cardboard or paper. Any set of
points in a 2D world (by definition) lies on the same 2D plane. Points on the
same plane are called coplanar. A plane is identified in the 3D world by a point
on the plane and a vector perpendicular to the plane. This vector is referred to as
the normal vector to the plane. For example, the x- and z-axes define a plane (the
x-z plane) with the origin as a point on the plane and the y-axis as the normal vector
to this plane. Similarly, the x-zand the y-z axes define planes. Can you identify the
normals for these planes? Alternatively two vectors in space (defined by three
points in space PlyP2 and P3) U = (P3 - P1) and V=( P2 - P I ) also define a plane.
The cross product of the two vectors U and V defines a third vector, which is
orthogonal to the plane defined by U and V. Mathematically the cross product of
two vectors is defined as follows:

If u = u2 v =
u3
Then the resultant cross product of the two vectors is

ulv2 - u2vl
This resultant vector is the normal vector to the plane formed by U and V as
shown in Fig.5.6.We shall use the concept of normal vectors heavily when we
learn about lighting in Chapter 6. The code for calculating the cross product of
two vectors and normalizing the result (creating a unit vector) can be found in
under the installed directory for our source code in the file: uti1s.h. The code is
shown below. We assume the vector values are stored in an array of 3 elements.

void crossproductlGLfloat *u, GLfloat *v, GLfloat *uv) {

Fig.5.6: Cross product of 2 vectors

In Example.52, we define three points in space:
GLfloat PI131
GLfloat P2[31
-- {0.5,0.,0.5);
{0.,0.,-0.5);
GLfloat P3[31 - {-0.5,0.,0.5);
The vectors, U and V, formed by these three points are defined as

where vecsub is a utility function defined in uti1s.h to calculate the difference

between two points. The cross product of U and V is the vector UV.

which is the (unit) normal vector to the plane defined. (Remember, this vector
just specifies the direction of interest - can you guess what the vector is?). We
draw the three coplanar points, and place a fourth point it's normal vector:

An easy way to define vertices is by making a call to the OpenGl function

The function expects a pointer to an array of three floating point values, as shown
in the sample code above. We then draw line primitives connecting all the points:

The result displayed forms a triangular pyramid as shown in Fig.5.7.Yes -this is our very fitst
3D model. That was hqwasn't it? Let us look deeper into the process of creating 3D models.
5.2 3 0 Modeling
CG scenes depict objects of all different kinds: people, trees, flowers, fire, etc.
Each one has radically diffcrcnt characteristics in terms of look and feel. So it
should come as no surprise that there is no one standard modeling
methodology-each object begs a different treatment to do full justice to the
model.
Thc most popular and commonly used method to model objects is to use
small polygons to approximate the surface. This method is popular due to ease
of use, speed of display, and the abundance of algorithms to deal efficiently with
polygon bascd models. Let us look into more detail on polygons and how thcy
are used to model more complex shapes. In Chapter 7, we shall see some
advanced techniques of representing models.

The Polygon
A polygon is much like a cutout piece of cardboard. Drawing a polygon is like
playing a game of connect the dots: each dot is a vertex defining the polygon.
You need at least three vertices to define the polygon. Each line of the polygon
is called an edge.
More formally, a polygon can be defined as a set of non-crossing straight
lines joining coplanar points to enclose a single convex area. Typically, most
graphics packages (including openGL) support single convex area polygons. A
single area means that the enclosed area should not bc divided. The convcx
requirement means that given any two points within the polygon, you must be
able to draw a straight line connecting these two points without going outside thc
area of the polygon, as shown in Fig.5.8.
The polygons arc required to be flat (that is, to be defined in a single plane).

(a) no area-illegal (b) single convex area (c) two areas, illegal

(d) single convex area (e) single, but non-convex area - illegal

Fig.5.8: Legal and Illegal Polygons

We saw earlier that three points in space define a plane. By definition, three
points also define a specific kind of polygon - the triangle. Hence, most graphics
software use triangles and triangle strips to define polygonal based models.
However, this approach is not essential, as we shall see when we define our
primitive objects.

Front-Facing and Back-Facing Polygons

Just as a piece of cardboard has two sides to it, a polygon has two sides or faces,
referred to as the front-face and the back-face. In CG, the order in which you
specify the vertices of the polygon defines the orientation of the polygon. By
convention, when polygon vertices are defined in counterclockwise order on the
screen the polygon is said to be front facing. If you reverse the order of the
vertices, the polygon is back facing: that is its back face is facing the viewer. If
you turn a front facing polygon around, it will become back-facing. The face of
the polygon is used extensively for defining different treatments of the two faces.

Fig. 5.9: Front- and back-facing Polygons

In Example 5-3, we connect the vertices of our pyramid to define triangle primitives.
We use GL-POLYGON as the geometric primitive passed to glBegin, with three points
defined in each block. We take care to define the polygons such that when the pyramid
turns, the vertices of the face facing the camera are in counterclockwise order.

void myPyramidO{
11 front face, vertices defined in counterclockwiie order
glColor3f (1.0, 0.0, 0.0);
glBeginlGL-POLYGON);
gWertex3fv(PII;
gRlertex3fv(wI;
glVertex3fvlP3I;
glEnd0;
I1 left face, vertices defined so that when this face is facing the camera, the points are in
counter-clockwise order
glColor3f 10.0, 1.Or 0.01;
glBegin(GL-POLYGON);
gNertex3fv(P1I;
gRlertex3fv(P2I;
gRlertex3fvluvI;
 glEnd0;
//right face
glColor3f 10.0, 0.0, 1.0);
glBegin(GL-POLYGON);
gNertex3fvlP3);
gNertex3fvlwl;
gNertex3fvlPZl;
glEnd0;
//bottom face
glColor3f (1.Or 0.0,I.O);
glBegin(GL-POLYGON];
glVertex3fvIP31;
glVertex3fvlP2);
glVertex3fvlPlI;
glEnd(1;
1
This function defines four polygons grouped together to form the shape of a
pyramid. The polygons are colored differently for sake of identification. By
default, OpenGL draws polygons as filled (with the current color).
The openGL function

can be used to define the treatment for each face of the polygons drawn.
We change the drawing mode for polygons so that we see the front faces of
our pyramid as filled, and the back faces as lines.:

The entire code can be found under Example.5-3/Example5-3.cpp.When you

run the program, only the front face is filled with the color red. The other faces
are drawn as lines, but you can still see them.

Back-face Culling
In real life, you cannot see all sides of a solid object. You see only the sides facing
you! The sides facing away from you are obstructed from view. To achieve this
effect in CG, we make use of the front- and back-facing polygon concept. (And
you were wondering why it was such a big deal!) The front- facing polygons are
facing towards the camera and drawn into the scene. The back-facing polygons
are facing away from the camera and hence are culled from the scene. This
process is known as back-face culling.
In Fig.5.10, we show our pyramid and a sphere with all the sides visible, and
 5. 3D MODEIING
CHAPTER

Fig.S.lO: Back-face culling applied to a) pyramid, and b) sphere

then the same objects with the back faces culled from the scene, thereby
generating a much more realistic image. To instruct OpenGL to discard back-
facing polygons, use the command

You will need to enable face culling with glEnable(). If we modify the init
function in Example5-3 to be defined as

void init(void){
llset the clear color to be black
glClearColor~O.O,O.O,O.OIO.O~;
glEnable(GL-CULL-FACE);
glCullFacelGL BACK);
g l ~ o l y g o n ~ o d eFRONT,
( ~ ~ GL LINE);
~ l ~ l ~ o l ~ ~ o n G~LINEI; ~ o d e l ~ ~ - ~ ~ ~ ~ ~
...
1
When you run the program, you will see only the pyramid's front facing
polygon. The back facing polygons are culled out--exactly what we wanted. The
reader is encouraged to try rotating the pyramid to see how back-face culling
affects the final output. By definition, as the pyramid turns around, the faces'that
were back-facing become front-facing (and vice versa).
Normal vectors can also be used to determine the orientation of polygons.
There is a one-to-one correspondence between the direction of the normal vector
and the order in which the polygon vertices are defined. If the normal vector of
the polygon points in the direction of the camera, then the polygon is front
 Fig. 5.11: Normals define the face of a polygon
facing, else it is back facing, as shown in Fig.5.11. We shall look into normal
vectors in more detail in the next chapter.

Polygon mesh
Polygons have limited usefulness by themselves. The real power of polygons
comes when they are used together in a mesh structure to make a more complex
shape. A polygon mesh (referred to as strips in openGL) is a collection of polygons
such that each edge is shared by at most two polygons, as shown in Fig.5.12.

-
Fig.5.12: Polygon mesh approximating a circle

In Example 5-3, we saw a pyramid defined by triangular polygons. Since the

triangles actually share their edges, we can think of the polygons forming a solid
model defined by the triangular mesh. (Ideally, triangles share edges, so the
shared edges need not be redefined. For the sake of simplicity, we will let the
triangles be defined individually. Since they are all connected, they effectively
define a mesh.)
In fact, the outer hull of all models can be approximated by a polygonal
mesh. The representation of models with just this mesh displayed is referred to
as wireframe representation. Almost all designers start off by constructing
wireframe representations of their models. Once they are satisfied with the hull
of their model, they will then add lighting and material properties to render the
models as a solid.
Let us look into how we can define primitive shapes using polygonal meshes.
5.3 3 0 Modeling Primitive Shapes
Primitive shapes have a well-defined mathematical formula that defines the
shape of their outer hull. Objects such as spheres, cubes, cones, etc. fall into this
category. These surfaces are also called implicit surfaces, since they are defined
by an implicit formula. Polygon meshes can be easily used to construct these
shapes. For the sake of convenience, most graphics software provide these
shapes as primitive objects in their own right. Let us look into some of these
shapes in more detail.

Cube
The cube model (or box) is one of the most common shapes occurring in nature.
If we know the center about which the cube is based, then the length, height and
depth completely define the cube as shown.

z
Fig.5.13: A cube defined by its three dimensions
In Example 5-4, we draw a cube, centered at the origin, as follows:

void Mycubel GLfloat length, GLfloat height, GLfloat depth){

11base
glBeginlGL-POLYGON);
glVertex3fl-lengthl2, -heightl2, depthl2);
glVertex3fl-lengthl2, -heightl2, -depthl2);
glVertex3fllengthl2, -heightl2, -depthl2);
gNertex3fllengthl2, -heightl2, depthl2);
glEnd0;

11front face
glBeginlGL-POLYGON);
glVertex3fl-lengthl2, -heightl2, depth/ZJ;
glVertex3fllengthl2, -height/2, depthl2);
glVertex3fllengthl2, heightl2, depthl2);
glVertex3fl-lengthl2, heightl2 depthl2);
glEnd0;
 11right side face
glBeginIGL-POLYGON);
glVertex3fllengthl2, -heightl2, depthl2);
glVertex3f(lengthl2, -heightl2, -depth/2);
glVertex3fllengthl2, heightl2 -depthl2);
glVertex3fllengthl2, heightl2, depthl2);
glEnd0;
11back side face
glBegin(GL-POLYGON);
glVertex3f(-lengthl2, -heightl2, -depthl21;
glVertex3fl-lengthl2, heightl2, -depthl2);
glVertex3f(lengthl2, heightl2, -depthl2);
glVertex3fllengthl2, -heightl2, -depthl21;
glEnd0;
I1left side face
glBeginiGL-POLYGON);
glVertex3fl-lengthl2, -height/2, depthl2);
glVertex3fl-lengthl2, heightl2, depthl2);
glVertex3fl-lengthl2, heightl2, -depthl2);
glVertex3f(-lengthI2, -height/Z, -deptN2);
glEnd0;

11top
glBeginlGL-POLYGON);
glVertex3f(-lengthl2, heightl2, depthl21;
glVertex3fllengtN2, heightl2, depthl21;
glVertex3f(lengthl2, heightl2, -depthl21;
glVertex3f(-lengthl2, heightl2, -depthl2);
glEnd0;

The built-in GLUT function

also generates a cube using polygons. It accepts the length of the cube as input
and draws a cube with equal sides, centered at the origin. The Display routine in
Example5_4/Exarnple 5-4.cpp makes calls to both cube drawing routines,
translating both of them for easier viewing:

void Display(void1
Sphere
Spheres are the most symmetrical shape occuring in nature. For this reason, they
are also the simplest to define. To completely define a sphere, we need to define
its radius, R, and the location of its center, 0. If we center the sphere around our
world origin, then any point P on the surface of the sphere can be defined by the
angles 8 and (I that it subtends at the origin as shown in Fig.5.14.
P=(Rcosecos$, Rsine, Rcosesin$)

Fig.5.14: Approximating a sphere

If we incrementally advance angle 8 from -90 to 90 degrees (-PI to PI radians)
and $ from 0 to 360 degrees (0 to 2PI radians), we can define a set of vertex
points approximating the surface of the sphere. These vertex points can then be
connected into a polygonal mesh to define the outer hull of the sphere. The closer
the intervals, the more accurate our model will be. The following shows the
subroutine to calculate the polygons of the sphere:

void MySphere(GLf1oat radius){

-
GLdouble inc P1112;
GLdouble theta, phi;
-
boo1 even true;

-
for (phi 0; phi < 2*Pl;phi+ -incl {
-
for Itheta --Pl12; thetaclP112-inc1;theta + inc){

glBegin(GL-POLYGON);
glVertex3f(radius*cos(thetal*cos(phi), radius*sinltheta),radius*cos(theta)"sin(phill;
glVertex3f(radius*cos(theta+ inc)*coslphi), radius*sin(theta + inc),radius*cosltheta + inc)*sin(phill;
glVertex3flradius*cos(theta + inc)*cos(phi+ incl,
radius*sin(theta+ inc),radius*cosltheta + incl*sin(phi + incl);
glVertex3flradius*cosltheta)*cos(phi+ incl, radius*sinlthetal,radius*cosltheta)*sinlphi + inc));
glEnd0;

The inc variable defines how many polygon slices and stacks we define for
the mesh defining the sphere. You can also use the GLUT built-in function:

which accepts the radius and the number of slices and stacks desired as arguments.
In Example5-5/Example5-5.cpp, we use both calls to generate spheres. Try
turning back-face culling on in this example by making calls to the functions

You will not see the back polygons in the sphere that we defined Yes, we did take care to
make sure polygons were definedwith the correct orientation.In the case of objects with a large
number of polygons, back-itice culling makes a big difference in the clarity of the image.

Cone
A cone can be thought of as having a circular base and a conical surface. The
 Fig.5.15: Approximating a cone
base of the cone can be defined as a circle on the x-z plane: the math is similar
to the way we constructed a circle in Chapter 1. We approximate the conical
surface of the cone by drawing a set of lines connecting the tip of the cone (a
point along the y-axis) to the vertices along the circular base. The following code
shows how to draw a cone (upper conic only) in a 3D world using openGL. The
cone is assumed to be centered about the origin.

void MyConelGLfloat radius, GLfloat height){

GLint i;
GLdouble theta, ntheta;
-
for (i 0; i < circlegoints; i + + 1 {
glBeginlGL-POLYGON);
--
theta (2*Pl*ilcirclegointsl;
ntheta (2*Pl*li + I)lcirclegoints);
glVertex3f(radius*coslthetal, 0, radius*sinltheta)l;
gNertex3f(OI height, 0);
glVertex3flradius*coslntheta), 0, radius*sinlntheta)l;

In Example 5-6, draw this cone, as well as the glut primitive:

The entire code can be found in Example56hple5-6.cpp. Some other primitive

shapes that glut provides support for are the disk, isohedron, and teapot. The teapot was first
modeled in 1975using curve and surface modeling techniques. It has now become an icon
for C q and hence is included in the primitives list. The call to generate a teapot in glut is
 Fig.5.16: The famous Utah teapot
Fig5.16 shows the teapot image generated using this hnction.

5.4 3 0 Modeling: Generic Shapes

So far, we have been discussing primitive objects that have a well-defined
mathematical formula to define their hull. What about objects that are more
generic in their representation? Surely, we cannot hope to find equations to
represent every possible object that occurs in nature.
For any kind of generic shape, we can always define a polygonal mesh with
sufficient complexity to approximate its hull. Chapter 7 details how Nurb surfaces
can be used to approximate hulls with greater accuracy and efficiency. The reader
is encouraged to imagine some generic 3D models and to try modeling them using
polygons, as discussed. It's not an easy task! How would you go about defining
the 3D vertices to model a tree, for example? Defining the vertices of the
polygons and/or curves in a 3D space can be a nerve racking experience if you
don't have the necessary mathematical tools. For this reason, a plethora of
modeling tools exist that will allow you to model your objects in 3D. We shall talk
more about Maya, the most widely used modeling tool, in Chapter 8.
Just like image information, model information can be saved in files. For this
chapter, we make use of model files to illustrate how generic objects can be
modeled using polygons.

Model files
What is stored in a model file? If the model is a primitive, then the key
parameters defining the primitive are saved. If it is a generic model, then
information about the (polygonal or cubic) primitives used to construct it are
saved.
There are many formats in which the model information can be stored. Maya
stores models in its own propriety format. For this book, we shall use an open
format called the VRML format. You can find many VRML models available for
free over the Internet. Indeed, we encourage you to go ahead and download
VRML model files yourself to experiment with. VRML files store polygonal
information as follows:
For each model in the file:
First, all the coordinates of the vertex points are listed as (x,xz) triplets. Then the
index of each vertex in this list is used to spec@ the vertices forming the polygonal
faces of the model, with a -1 as the delimiter. For example, to draw two triangles
that form the faces of a unit square (Fig.5.17), the vertices would be listed as

0. 0.0. I1vertex 0
1.0. O., I1vertex 1
1. 1. O., //vertex2
0. 1.0. Ilvertex 3

and the indices forming the faces would be listed as

0,1,2,-1 11bottom triangle

2,3,0,-1 11top triangle

Fig.5.17: Triangular faces forming a square

Normal vectors to each polygonlvertex are defined in a similar manner. First the
normal vectors are defined, and then the indices are defined determining the face
to which the normal applies. Of course, we can actually calculate the normal
vectors for each polygon, by computing the cross product of its defining vectors!
VRML models also contain material information, which we shall not look into.
A version of code to read in a very simple VRML file can be found under the
directory where you installed our software, in the files: vrml.cpp and vrm1.h. In
these files, we define a function

int ReadVRMLkonst char "filename,

GLfloat "coordinates, GLfloat "normals,
GLint "indices, GLint "nindices,
GLint "noofpoly, int MAXSHAPES, int MAXCOORDS
I;
that will read the VRML file, and fill in the coordinate and normal vertex data.
indices and nindices array hold the index values of the vertices which define the
polygons and normals for this shape.
noo&oly array holds a count of the number of polygons read for each shape
in the file.
M S H A P E S determines the maximum number of shapes that will be read
from the file.
M C O O R D S indicates the maximum number of coordinates that each
shape may have.
In Example 5-7, we read in the VRML file robot.wrl: a model of an Android
model. The model itself consists of different shapes forming the head, legs,
hands, etc of the Android. The code to read in the model is simple. We define the
MAXSHAPES that we can read in to be 25 and the maximum coordinate data
per shape as 3000:

#define MAXSHAPES 25
#define MAXCOORDS 3000;

The array to hold the coordinate data is defined as

This structure will hold the coordinate data for up to 3000 vertices (900013)per
shape and upto 25 shapes. The normal data is similarly defined, although we will
not use it in this chapter. The index array holds information for up to 3000
triangular faces and up to 25 shapes.

GLint indices[MAXSHAPES1[3*MAXCOORDSl;

For each shape, both arrays hold the three coordinates/indices of each
vertedpolygon in a linear one-dimensional manner. We maintain a count of the
number of polygons defined by each shape

GLint noofpoly[251;

In the main code, we read in an android model as follows:

-
noofshapes ReadVRMLI'..\\Models\\robot.wrl", &coords[OJ[O], &normals[Ol[Ol,
&indices[Ol[Ol,&nindices[Ol[OII
&(noofpoly[Ol), 25,3000);

We can draw the models in many different ways. The simplest way is to draw the
triangular faces one by one using the glVertex command.
for (j-O;j<noofshapes;j+ + 1 {
-
for (i O;i<noofpoly[j1"3;i -i +3) {
glBeginlGL-TRIANGLES);
glVertex3f~coords[jl[3*indices~[ill,
coords[jl[3*indices[jl[il+ 11,
coords[jl[3*indices[jl[il+ 21);
glVertex3flcoords[jl[3*indices[jl[i + 1I],
coords[jl[3*indices[jl[i + 1I+ 11,
coords[jl[3*indices[jl[i + 1I+ 211;
glVertex3f(coords[j1[3*indices[jl[i+ 211,
coords[jl[3*indices[jl[i + 21+ 1I,
coords[jl[3*indices[jl[i + 21+ 211;
glEnd0;
11
An easier way to define the vertices is by making a call to the OpenGl function

Using this function, the display code can be modified to be

-
for (j O;j<noofshapes;j + +1 {

Fig.5.18: An Android, read from a VRML model

 -
for (i- O;icnoofpoly[j]*3;i i+3l {
glBegin(GL-TRIANGLES);
glVertex3fvl&~coords[j][3*indices~[i]]ll;
glVertex3fv(&(coords~[3*indices[jl[i+ 1]Ill;
glVertex3fvl&lcoords~[3*indicesl,jl[i+ 21111;
glEnd0;
11
Fig.5.18 shows the output of the code. The entire code can be found in
Example5-7/Example5-7.cpp,and the VRML file, robot.wr1, can be found under
the Models directory under the installed folder for our sample code. You will
need to compile and link the program with the files: vrm1.h and vrml.cpp.
Readers are encouraged to experiment with their own VRML models. You can
download these models for free from the Internet. You may have some trouble
with world coordinates of differing objects. We discuss how to modify the world
coordinate space in 3D in the next section.

5.5 3 0 Transformations
We have seen how to create models in the 3D world. We discussed transforms in
Chapter 2, and we have used some transformations to view our 3D models
effectively. Let us look into 3D transformations in more detail.
Just as 2D transformations can be represented by 3 x 3 matrices using
homogenous coordinates, 3D transformations can be represented by 4 x 4
matrices. 3D homogenous coordinates are represented as a quadruplet (x,y,z, W).
We discuss the representation of the three most commonly used transformations:
translation, scaling, and rotation. 1 O O T x
The translation matrix T(Tx, lj,Tz) is defined as T =

so any vertex point P =

Scaling is similarly extended to be defined as

r s x o o o 1
o s y o o
o o s z o
 Recall from Chapter 2 that scaling occurs about a point. This point is called
the pivotpoint. The scaling matrix defined above scales about the origin. You can
define models to have different pivot points by setting the transformation stack
appropriately. We shall see in greater detail, on how models can be defined to
have a predefined pivot point.
Rotation matrices are defined uniquely based on the axis of rotation and the
pivot point. The axis of rotation is a normalized vector defining the axis in 3D space
along which the rotation will occur. (A 3D vector will simply be the triplet (x,y,z)
defining the direction of the axis). The 2D rotation we saw earlier is a 3D rotation
about the z-axis (the vector (0,0,1)), and the world origin (0,0,0) and is defined as:
cos(0) -sin(0) 0

Similarly the rotation matrix about the x-axis is defined to be

and about the y-axis as:

Just as with 2D transforms, 3D transforms can be composed together to provide

desired effects. The order in which the transforms are applied is important.
Recall from our previous section that the OpenGL commands for matrix operations

void glLoadldenti(void1

sets the current matrix to be the 4 x 4 identity matrix.

glLoadMatrix4f(GLfloat *m)

sets the matrix passed in to be the current transformation matrix.

glMukMatrix4f(GMoat "m)

multiplies the current matrix by the matrix passed in as the argument.

 The argument for all these commands is a vector of 16 values (ml,
m2,..,m16) that specifies a matrix M as follows:

m9 m10 m l l m12
m13 m14 m15 m16 1
Additionally, OpenGL provides three utility functions for the most
commonly used transformations.

multiplies the current transformation matrix by the matrix to move an object by

(Tx,Ty,Tz) units

multiplies the current transformation matrix by a matrix that rotates an object in

a counterclockwise direction by theta degrees about the axis defined by the
vector (x,y,z)

multiplies the current transformation matrix by a matrix that scales an object by

the factors, (Sx,Sy,Sz).
All three commands are equivalent to producing an appropriate translation,
rotation, or scaling matrix and then calling glMultMatrixf0 with that matrix as
the argument. Let us apply some transformations to the famous pyramid that we
created earlier. In Example5-8, we rotate the pyramid about the y-axis and use
double buffering to display the rotation as an animation.

The entire code can be found in Example5-8/Example5-8.cpp. Readers

should experiment rotating about both the x- and z-axes to see what results they
get So now that we have covered some of the basics of 3D modeling, you are
probably asking: how are we defining the 3D world? What are its extents? How
is this world being mapped onto the 2D screen? Let us take a closer look at how
this process is accomplished.
In Chapter 2, we saw how to map a two-dimensional world onto the two
dimensional screen. The process was a simple mapping of 2D world coordinates
 onto the 2D viewport of the window. Model positions are first clipped against the
2D world coordinates and then mapped into the viewport of the window for
display.

5.6 Viewing in 3 0
The solution to transforming a 3D world into a 2D plane is a little more complex
and is accomplished by the use of projections. Projections provide the transforms
needed to map 3D points onto a 2D plane, as shown in Fig.5.19. To determine
the projection, we need to define a view volume of the world (the world
coordinates), a projection that defines the mapping of the view volume onto a
projection plane, and the viewport for final display. Conceptually, objects are
first clipped against the 3d view volume, projected, and then finally mapped into
the specified viewport.
Whew! That was a mouthful. Let us see if we can understand this process by
way of a real-world analogy. The essential goal in 3D CG is to give the viewer
the impression that he is looking at a photograph of a three-dimensional scene,
much the same way we photograph our 3D world onto a 2D film. 3DCG
simulates the process of the real-world camera to create 3d images. Let us first
understand how a real camera records an image of our real live 3D world by
using the example of a simple pinhole camera. We will then extend this
discussion to see how CG simulates this process.

Fig.5.19: Projecting a 3D object onto a plane

Pinhole Camera
Fig.5.20 shows a pinhole camera. The camera is a simple box with photographic
film F at the back and a small hole H that allows light in when opened.
Consider what happens when we take a photograph of object X. We open
hole H for a fraction of a second. A ray of light hits X at point P, passes through
H and hits the film at P', causing the film to be exposed at this point. The point
P' forms the image of point P on the negative. All points on the object are
mapped onto the film in this fashion. Another object Y, of the same size but
 ObjectX

Fig.5.20: Pinhole Camera

further away, gets mapped on as Y' and is actually smaller in size. Objects such as
Z, whose intersecting rays of light do not pass through H, are not photographed.
Notice that the image formed is inverted! The classic computer graphics version of
the camera places the film in front of the pinhole. This placement ensures among
other things, that the image is right side up. We rename the pinhole as the eye,
viewpoint, or cameraposition and the film as the imageplane, as shown in Fig.5.21.
The process of filming the scene remains very much the same. Rays (called
projectors) are generated from the viewpoint to the objects. Objects for which
these rays intersect the image plane will be processed and displayed on screen.
Others such as Object Z, will be clipped from the image.
The image plane itself can be thought of as being composed of a two-
dimensional grid containing color information that is then mapped onto the
actual screen pixels by the viewport transformation. For the sake of convenience,
we can think of the image plane itself as being composed of pixels.

Fig.5.21: Computer Camera

The CG Camera
The steps needed to generate the CG image can be defined analogously to the
steps we would take a photograph in the real world. We need to define the
following:
1. Viewing Transformation
The viewing transformation sets the viewing position and is
analogous to positioning and aiming a camera.
 2. ObjectIModeling Transformation
The modeling transformations are used to position and orient the
model within the world. This process is akin to arranging the scene to
be photographed into the desired composition. For example, you can
rotate, translate, or scale the model-or perform some combination of
these operations. We saw how to do this in a 2D world. In 3D, the
concept is enhanced to include a third dimension defined by the z-axis.

Fig.5.22: Modeling and Viewing Transformation

3. Projection Transformation
Specifying a projection transformation is akin to choosing a
camera lens or adjusting the zoom. This transformation determines
the extents of the world in view, also called the viewing volume.
In addition, the projection transformation also determines the
location and orientation of the image plane and how objects in the
scene are clipped and projected onto this plane. This process
determines the world coordinates of the scene.

viewing volume
Fig.5.23: Projection Transformation

4. Viewport transformation
The viewport transformation determines the shape of the available
screen area into which the scene is mapped and finally displayed.
This process is analogous to determining how large the final
 photograph will be. This step is same as in the 2D world - when
we finally map the world coordinates onto the display.

Fig.5.24: Viewport Transformation

After these steps are performed, the picture can be snapped, i.e., the scene
can be drawn onto the screen.
Each transformation step can be represented by a 4 x 4 matrix M (as opposed
to 3 x 3 for a 2D world). The coordinates of each vertex point P (in object space)
is successively multiplied by the appropriate transformation matrix to achieve
the final position P' (in world space) as shown in Fig.5.25.

object coordinates wind o w coo rdinates

Fig.5.25: The transformation sequence
Each transformation step is saved in its own matrix stack. We saw the OpenGl
command to specify the matrix stack that you wish to operate on is

glMatrixMode (int model

where mode is one of GL-MODELVIEW, GL-PROJECTION, or GL-TEXTURE.

Subsequent transformation commands affect the specified matrix stack. We
saw in earlier examples how to set the ModelView matrix stack and modify the
transforms applied to the models in the scene. We also saw how viewport
transformations are accomplished by the OpenGL command, glviewport.
Note that only one stack can be modified at a time. By default, the
ModelView stack is the one that's modifiable.
Let us look at each transformation stack one at a time. We begin with model
transformations since we have already seen it before, and it's an intuitive place
to begin our discussions.

ObjectIModel Transformations
We can move, resize, or orient the models in our 3D world differently to create
a composition or scene. We saw in Chapter 2, that these transformations are
called object transformations or model transformations. There is an alternative
way to think about model transformations: moving the camera in one direction
will have the same effect as moving all the models in the scene in the opposite
direction! For this reason, viewing and modeling transformations are
inextricably related in OpenGL and are in fact combined into a single modelview
matrix. You should find a natural approach for your particular application that
makes it easier to visualize the necessary transformations, and then write the
corresponding code to specify the matrix manipulations.
You must call glMatrixMode0 with GL-MODELVIEW as its argument prior
to performing modeling or viewing transformations. In Example 5-9, we modify
the Modelview matrix to move our pyramid model back in z- by 4 units, scale it
up and constantly keep rotating it around the y-axis.

glloadldentity 0;
11 translate pyramid back by 4 units, and down by 1 unit
gITranslatef(O.,-I .,-4);
/I scale pyramid
glScalef(2.,2.,1 .I;

11 rotate the pyramid

glRotateflang,O.,l .,O.I;

if (ang > -360) ang - 0; else ang+ +;

11 draw the pyramid
myPyramid0;
glFlush0;
glutPostRedisplay0;

Which transform is being applied first? What will happen if we change the order
of the transforms? The entire code can be found in Example59/Example5-9.cpp.

Viewing Transformation
A viewing transformation changes the position and orientation of the viewpoint
itself. Essentially, the viewing transformation repositions the camera and the
direction it is looking. As discussed earlier, to achieve a certain scene
composition in the final image or photograph, you can either move the camera
or move all the objects in the opposite direction. In Example.5-9, the translation
routine can be thought of as actually moving the camera position by 4 units
towards the object, instead of the object being moved back. Readers are
encouraged to try experimenting with this approach of transforming the scene.
Remember that transformations are applied in the reverse order of the
commands specified. Viewing transformation commands must be called before
any modeling transformations are performed. This will ensure that the viewing
transformations are applied to all the models after the model transforms.
You can manufacture a viewing transformation in OpenGL using different
techniques. One is using the standard glTranslate and glRotate functions. A more
popular method is to use the Utility Library routine gluLookAt() to define a line
of sight.

void gluLookAt(GLdoubleeyex, GLdouble eyey, GLdouble eyez, GLdouble centerx, GLdouble centery,
GLdouble centen, GLdouble upx, GLdouble upy, GLdouble upz);

This call defines a viewing matrix such that the desired viewpoint is located at
(eyex, eyey, eyez). The centerx, centery, and centerz arguments specifies a point
in the scene that is being looked at. The upx, upy, and upz arguments indicate
which direction is up (that is, the direction from the bottom to the top of the
viewing volume). This matrix is then multiplied to the current modelview matrix.
In the default position, the camera is at the origin, is looking down the negative
z-axis, and has the positive y-axis as straight up. This is the same as calling

Fig.5.26: Camera position using gluLookAt

The following code from ExampleS-I O/Example.5-I 0.cpp moves the viewpoint
closer to and farther from the pyramid, as the pyramid continues its rotation
about the y-axis. Notice that we call the viewing transform before calling the
object transforms. The matrix stack is specified in the init() fbnction.

- -
GLfloat zdist 3., zmove I;
void Displaylvoidl
{
glutSwapBuffers0;
glClear (GL-COLOR-BUFFER-BIT);
glLoadldenti I);

gluLookAt(O.,O.,eyez. O,O,-100,0.,1 .,O.);

-
if (eyez >20l zmove -1;
if (eyez <3) zmove 1;
eyez+ -move*0.2;
-
11 rotate the pyramid
glRotateflang,O.,l.,O.l;
- -
if lang > 360) ang 0; else ang + + ;
11 draw the pyramid
myPyramid0;
glFlush0;
glutPostRedisplay0;
1
How would you define the transforms such that the camera rotates around the
pyramid? Hint: You can use the equations we derived for points along a circle.

Projection Transformations
In general, projections transform points in an n-dimension coordinate system to
a coordinate system with fewer than n-dimensions. In 3DCG we are interested in
the projection from a 3D space to a 2D space--just like a shadow is a projection
of a 3D object (you) onto a 2D plane (the ground).
The projection transformation is defined by a viewing volume (the world
coordinates), the camera position or the viewpoint, and the image plane, also
referred to as the projection plane. The class of projections we deal with here is
known as planar geometric projections because the projection is onto a plane
rather than some curved surface. The Omnimax film format, for example, uses a
non-planar projection plane.
The viewing volume and the viewpoint determine how the 3D object is
projected onto the plane. We use straight lines, called projectors. Projectors are
drawn from the center of projection (the viewpoint) to each vertex point on the
object. The points of intersection between the projectors and the image plane
forms the projected image. The two basic classes of planar projections are
perspective and parallel, as shown in Fig.5.27.
 Center of Center of projectibn .."
projection at infinity
Fig.5.27: Perspective and Parallel Projection

Perspective Projection
The most unmistakable characteristic of perspective projection is perspective
foreshortening: the farther an object is from the camera, the smaller it appears in
the final image. This effect occurs because the viewing volume for a perspective
projection is a frustum of a pyramid (a truncated pyramid whose top has been cut
off by a plane parallel to its base). Objects that are closer to the viewpoint appear
larger because they occupy a proportionally larger amount of the viewing volume
than those that are farther away. This method of projection is commonly used for
animation, visual simulation, and any other applications that strive for some degree
of realism because the process is similar to how our eye (or a camera) works. The
following command defines the frustrum of the viewing volume:

void glFrustum(GLdouble left, GLdouble right, GLdouble bottom, GLdouble top, GLdouble
near, GLdouble far);

This function calculates a matrix that accomplishes perspective projection and

multiplies the current projection matrix by it. The viewing volume is defined by
the four sides of the frustum, its top, and its bottom, which correspond to the six
clipping planes of the viewing volume, as shown in Fig.5.28. Essentially, the
viewing volume defines the world coordinates within which our 3D objects reside.
Objects or parts of objects outside this volume are clipped from the final image.
viewing volume

4
far b

Fig.5.28: Viewing Frustrum

 The projection plane is defined as a plane parallel to the base of the pyramid
and centered at the viewpoint position. The following code from Example5-11,
shows the effect of changing the frustrum parameters. Changing the left, right,
bottom and top planes is like changing the viewing lens on a camera - you can
view your scene with a wide angle or a telephoto lens!

glMatrixMode IGL-PROJECTIONI;
glLoadldentity0;
glFrustum (-1.O*lens, I .O*lens, -1.O*lens, I .O*lens, 1., 20);
11 gluPerspectiie(fang, aspect, 1., 20.01;
glMatrixMode (GL-MODELVIEW);
glLoadldentity(1;
gluL00kAt(0.,0.,3, O,O,-100,0.,1.,0.1;
11 rotate the pyramid
glRotatef(ang,O.,l.,O.);
-
if (ang > 360) ang
I1 draw the pyramid
- 0; else ang + + ;

myPyramidll;
glFlush(l;
glutPostRedisplay0;

If you change the near and far clipping planes instead, you will see the
pyramid gradually being clipped out of the scene. The glFrustum command is
not always intuitive. An easier command to define a perspective projection is

void gluPerspective(GLdoublefovy, GLdouble aspect,

GLdouble near, GLdouble fad;

fovy is the angle of the field of view in the x-z plane (shown in Fig.5.28) This
value must be in the range [0.0,180.0].
aspect is the aspect ratio of the frustum, that is its width divided by its height.
Example5-11 can be modified to replace the command glFrustrum with the
command:

gluPerspectivelfang, aspect, 1., 20.01;

where the field of view (fang) of the camera changes from 30 degrees to 160
degrees to achieve the same effect as seen by glFrustrum. In general, we always use
the gluperspective function to define the perspective projection in our programs.

Parallel Projection (Orthographic)

Parallel projections are projections whose center of projection is at infinity.
These projections ignore the location of the current viewpoint in their
calculations. The projectors used are parallel lines.
 't top view

side view
/ front view
Fig.5.29: Orthographic projection: viewing volume
The output of a parallel projection is not realistic because objects retain their
size no matter where they are located. It is very usefbl for calculating
measurements and sizes of objects and is often used when modeling objects, in
engineering drawing, and in blue prints of architectural plans.
Parallel projections are called orthographic if the direction of projection is
orthogonal (at 90 degrees) to the projection plane. The most common type of
ortho projections are along the x-, y- and z- axis (called the front, top and side
views) where the projection planes are on they-z, x-z and x-y planes respectively
as shown in Fig.5.29. These projections are used extensively in modeling tools
to enable the process of 3d modeling.
With an orthographic projection, the viewing volume (the world coordinates)
is a box, as shown in Fig.5.30

near far
bottom
Fig.5.30: Orthographic viewing volume

Unlike perspective projection, the size of the viewing volume doesn't change
from one end to the other, so distance from the camera doesn't affect how large
an object appears to be. The direction of the viewpoint defines the direction of
 5 3 0 MODEIING
CHAPTER 117

projection, but its location is ignored. The openGL command to specify the view
volume for orthographic projections is defined as

void glOrtho(GLdouble left, GLdouble right,

GLdouble bottom, GLdouble top,
GLdouble near, GLdouble far);

We saw a specialized version of a 2D orthographic projection in the Chapter 2-

glOrtho24 which defined a 2D rectangle for the viewing area. In Example 5-12, we display
the pyramid using a 3D orthographic projection. Since we don't have any animation
defined, we can define the projections so that they are set only in the reshape hction:

void reshape lint w, int h)

{
11on reshape and on startup, keep the viewport to be the entire size of the window
glviewport 10, 0, IGLsiieil w, IGLsizei) h);
glMatrixMode IGL-PROJECTION);
glLoadldentity0;
I/ Set up the ortho view volume
glOrtho(-I,I,-I ,I,I.,201;
glMatrixMode IGL-MODELVIEW);
//Define viewing position as 5 units back in Z, and
gluLookAtl0.,0.,5, o,o,-100,0.,1 .,O.I;
I/ rotate for top view
IlglRotatefl90,l .,O.,O.);
11rotate for side view
IlglRotatefl-90,0.,1.,0.);

This code will display the front view of the pyramid. To see the top view, we
need to define the direction of projection to be along the y-axis. This is achieved
by rotating the viewing position 90 degrees about the x-axis by inserting the
following lines of code after the gluLookat command:

I1rotate for top view

glRotatef190,1.,0.,0.);

Similarly, rotate the viewing position by 90 degrees along the y-axis to get a side
view. The different-colored faced of the pyramid will provide you with a good
indication as to where you are looking.

Vie wpolf Transformations

We looked into details of the viewport transformation in Section 1. The viewport
 transformation transforms the image plane into the specified region of the
window.
By default the viewport is set to the entire pixel rectangle of the window that
is open. You can also use the glViewport0 command to also choose a smaller
drawing region; for example, you can subdivide the window to create a split-
screen effect for multiple views in the same window. Try experimenting with
different viewport settings within your own programs.
Transformations and matrix stacks are not only useful for defining the 3D-to-
2D mapping. They prove to be an invaluable tool when constructing models as
well. Let us look into this aspect of modeling before we end the chapter.

5.7 Hierarchical Modeling Using Transformations

Models are usually not constructed as monolithic blobs. They usually have a
hierarchical structure: smaller models called shapes/components are used as
building blocks to create higher-level entities, which in turn serve as building blocks
for higher-level entities, and so on. A hierarchy enables the creation of complex
objects in a modular fashion, typically by repetitive invocation of building blocks.
The model hierarchy can be symbolized as a tree, with the components defined as
nodes of the tree. The nodes at the top of the hierarchy are the parent objects. Parents
have links to their child objects, who have further links to their children etc.
The child objects are defined in a default orientation and must be transformed
in order to be correctly positioned with respect to their parent object. To
accomplish this, parent objects call each child objects in the hierarchy, passing
to it the appropriate geometric transforms corresponding to its scale, orientation
and position. The parameters are defined within the parent's transformation
space. This means that the children inherit their parent's transforms, so a link
ensures that the spatial position of the child gets updated when the parent is
transformed. Transforming the children has no affect on the parent.
The hierarchy of the model determines the motion definition for the model
and its components. Careful thought needs to be put into constructing a hierarchy
for a model. The best choice of the hierarchy is one that takes into account the
movement/animation of the model within scene.

Fig.5.31: A Snowman
 Consider building the model of a snowman. We shall refer to him as Snowy,
as shown in Fig.5.31
Let us first consider the component objects that build up the snowman. The
base, tummy, and head are of course all spheres. The hands could be modeled
using cones or cylinders and the eyes using disks. The carrot shaped nose is just
a simple cone.
How would we define the hierarchy of such a model? We definitely want the
eyes and the nose to ride along with the face. Therefore we would define them to be
children of the head object. Similarly, the hands can be grouped as children of the
tummy. Now the base, tummy and head need to be parented in a way that would
enable us to move them as a single unit. In order to do this, we define what we call
a Null Node. A null node is used solely for the purpose of grouping together
components in a hierarchy. It has no graphical display in our 3D world. It does have
a location, which defines the pivot for any rotation and scale transforms applied to
it (and hence to its children). The introduction of the null node would lead us to a
hierarchical structure of the snowman as shown in Fig.5.32. Let us see how we can
use matrix stacks to define the hierarchical model of the snowman using OpenGL.

Fig.5.32: Hierarchy of the snowman model

Matrix Stacks
The concept of a stack of matrices is used for constructing hierarchical models.
Matrices can be pushed on top of the stack and popped out by using the Push and
Pop matrix operators. At any given time, only the topmost matrix is used to
define the current transformation state. The stack being used is determined by
glMatrixMode0 - in this case we will just be using Modelview matrix stack.
OpenGl provides two commands to deal with stacks of matrixes.
The OpenGL command

void glPushMatrixlvoid1;

pushes all matrices in the current stack down one level. The topmost matrix (Mc
in the Fig.5.33) is copied, so its contents are duplicated in both the top and
second-from-the-top matrix. This is an effective way to save the current
transformation state while applying further transformations at the top.
 Fig.5.33: Push operator
The openGL command,

void giPopMatrix(voidl;

pops the top matrix off the stack, destroying the contents of the popped matrix.
What was the second-from-the-top matrix becomes the top matrix. From the
previous fig., we effectively pop back to the saved transformation state: matrix
Mc. If the stack contains a single matrix, calling glPopMatrix() generates an
error. The Push and Pop commands are very usefbl to 'remember the last
transformation and to pop back to it when needed

Romove
additional top most
transforms matrix
Pop back 1-
to the -4 Mc 1
previously
saved
state

Fig.5.34: Pop operator

The Snowman Model using Hierarchy

Hierarchical models are implemented using the Modelview matrix stack.
First, the drawing routines of the building blocks are defined. The routine
defines the blocks in a default position and orientation that are convenient for the
modeler: the object coordinate system or object space. For the snowman model,
the drawing routines for the building block objects-spheres, cones and disks-
are already defined in the GLUT library. The sphere and disk primitives have a
default position centered about the origin. The cone is based on the x-z plane,
centered about the origin.
 Within the drawing routines for each of our components, we first push the
current matrix stack to remember the state. We apply the transformations (passed
as parameters from the parent) for this component and finally pop back to our
original state before returning from the call.
Let us see how to implement the hierarchy of Snowy using OpenGL.
We define a one dimensional array of size = 9 to define the (Tx, Ty, Tz, Rx,
Ry, Rz, Sx , Sy, Sz) transformations (also referred to as xforms in CG lingo) for
each of the components of the snowman. The array is set to the identity
transformation to begin with.

//Top level snowman xforms

-
GLfloat snomanXforms[9] {0.,0.,0.,0.,0.,0.,1.,1.,1.);
llxform of bottom
-
GLfloat botXforms[9] {0.,0.,0.,0.,0.,0.,1.,1.,1.);
11xform of tummy
-
GLfloat tumXforms[9] {0.,0.,0.,0.,0.,0.,1.,1.,1.);
11xform of head

IlXform of eyes
-
GLfloat headXforms[91 {0.,0.,0.,0.,0.,0.,1 .,I.,I.);

--
GLfloat IEyeXforms[9] {0.,0.,0.,0.,0.,0.,1 .,I.,I.);
GLfloat rEyeXforms[9] {0.,0.,0.,0.,0.,0.,1.,1.,1.);
IlXform of nose
-
GLfloat noseXforms[9] {0.,0.,0.,0.,0.,0.,1.,1.,1.);
IlXform of hands
--
GLfloat IHandXforms[S] {0.,0.,0.,0.,0.,0.,1.,1.,1.);
GLfloat rHandXforms[91 {0.,0.,0.,0.,0.,0.,1.~1.,1.);

The drawing routine for each component accepts its corresponding xform array.
The drawing routine first saves the current transformation state by using a
PushMatrix comand. It then applies its local transforms, and then pops the stack
back to the original transformation state. For example, the bottom of the
snowman, with a radius of 1.5, units can be defined as follows:

void draw-Bottom(GLfloat *botXformsI{

glPushMatrix0; 11save state
lldefault state of bottom
glTranslatef(O.,l.5,0.); 11translate bottom up by 1.5 units
11apply local transforms to the bottom
glTranslateflbotXforms[Ol, botXforms[ll, botXforms[ZII;
glRotatef(botXforms[3], 1,0,0I;
glRotatef(botXforms[4], 0,1,0I;
glRotateflbotXforms[51,0,0,1 I;
glScaleflbotXforms[6], botXforms[7], botXforms[81);
I1 actual drawing code
 glutSolidSpherell.5,20,201;
glPopMatrix0; 11 pop back to original state
I
By default, we translate the base of the snowman up by 1.5 units, so its base
is resting at the origin. Transformations to the bottom component are applied by
modifying the botxforms array.
The pivot point for each component and its children is very important in a
hierarchical model. Each component has a pivot point based on the location of
the origin in its local object space. This point is also called the local origin.Local
transformations for components are applied using this pivot point. In this
example, the pivot point for the bottom is at the current location of the origin-
the center of the sphere. If we had moved the lines of code:

gITranslatef(0.,1.5,0.1; 11 translate bottom up by 1.5 units

to be defined just before the actual drawing code, then the local origin would be
located at the base of the sphere when the transforms are applied. The pivot point
for the bottom would hence be at its base.
This technique is often used to modify pivot points for components within a
model. In this case, we actually wanted to re-position the sphere 1.5 units up as
well. If we do not wish for the actual model to be re-positioned, we would then
apply a reverse translation to move the model back to its original location. This
process would modify the pivot point without affecting the position of the model.
We saw this process at work in Chapter 2, Example2.4.
All children of a component are drawn within the transformation state of the
parent. This ensures that the child is transformed along with its parent. Due to
the Pop operation, any transformations applied to the child are not applied back
to the parent. For example, the hands, which are children of the tummy, are
defined within the transformation stack of the tummy. The transforms of the
hands are applied only in the local routine defining the hand.

void draw-Tummy(GLfloat "tumXforms, GLfloat "IHandXforms, GLfloat "rHandXforms){

glPushMatrix0;
glTranslatef(tumXforms[Ol, tumXforms[ll, tumXforms[2]1;
glTranslatefl0.,3.9,0.1;
glRotatefItumXforms[31,1,0,01;
glRotatef(tumXforms[4], 0,1,01;
glRotatef~tumXforms[5110rO11 1;
glScalef(tumXforms[6], tumXforms[7], tumXforms[811;

{
glutSolidSphere(l,20,201;
I1 draw children within parent's transformation state
 The tummy has a radius of 1 unit and is transformed up by 3.9 units, so it
rests on top of the bottom (!). Notice that we apply the scale and rotate transform
before the translations. Doing so ensures that the pivot point for these operations
is at the local origin-the center of the sphere.
The left hand (and similarly the right hand) are defined as cones. By default,
they are rotated by 90 degrees and translated so that they are sticking out from
the surface of the tummy. The hands can be moved by modifying the
1HandsXforms array. Additionally, they will also be transformed when the
tummy is! The pivot point for the scale and rotate transforms for the hand is at
the local origin of the cone which is located at the center of it's base.

void draw-LHandlGLfloat *IHandXforms){

glPushMatrix0;
glTranslatef(-I.,0.,0.1;
glTranslatefllHandXforms[Ol, IHandXforms[ll, IHandXforms[211;
glRotatefIlHandXforms[31, 1,0,0);
glRotatefllHandXforms[4], 0,1,0);
glRotatef(lHandXforms[511010111;
glScalefIlHandXforms[61, IHandXforms[71, IHandXforms[811;
//default state is pointing outward
glRotatefl-90.,0,1,01;
{
glutSolidConel0.1,1.5,5,5~;
1

The head and its children (eyes and nose) are similarly defined. The final
snowman drawing code is as follows:

void draw-SnowManlGLfloat "snomanXforms, GLfloat *botXforms, GLfloat *turnXforms, GLfloat

*headXforms, GLfloat *Kforms, GLfloat *rXforms, GLfloat *noseXforms, GLfloat "IHandXforms,
GLfloat "rHandXformsl{
glPushMatrix0;
glColor3fll.,l .,I.I;
glTranslatef(snornanXforms[OI, snornanXforms[ll, snornanXforms[211;
glRotateflsnornanXforms[31,1,0,01;
glRotateflsnornanXforms[411011,01;
 I
draw Bottom(botXforms1;
draw~~umm~(tum~forrns, IHandXforms, rHandXforms1;
draw-HeadlheadXforms, Kforms, rxforms, nosexfoms);

Note that the snowman object itself has no drawing component to it. Since there
are no transforms before the scale and rotate calls, the pivot point is at the local
origin: which happens to be the base of the snowman. This pivot is used for
transforms applied to the entire snowman hierarchy (snowmanXforms).
In Example.5-13, we rock the snowman back and forth about its base by
rotating it along the z-axis. At the same time, we also swing its hands up and
down and nod its head left to right (rotation along the y-axis)

void Display(void1
{
GLfloat ypos, xpos;

---
snomanXfoms[5] + 1"sign;
IHandXforms[S] + 2*sign;
rHandXforms[5] + -2"sign;
- --
headXforms[4] + 2"sign;
if (snomanXforms[51 301
-
sign -1;
else if IsnomanXforms[5]
-
sign 1;
-- -301

glClear(GL-COLOR-BUFFER-BIT I GL-DEPTH-BUFFER-BIT);
draw~SnowManIsnomanXforms, botXforms, tumXforms, headxforms, IEyeXforms,
IEyeXforms, noseXforms, IHandXforms, rHandXforrns1;

Try playing around with the various transformations and verify that the
hierarchy works as intended. The entire code can be found under Example.5-13,
in files: Example5-13.cpp, Snowman.cpp and Snowman.h. Make sure you
completely understand the concept of pivot points and how they can be modified
before moving on to the next section. Now, consider a slightly more complex
example for modeling a hierarchical model, that of the android.

Right Thigh Left Thigh

Right Calf

Fig.5.35: Components of the Android and its hierarchy

The entire Android is rooted at the Root component.

The lower torso is rooted at the pelvis, about which swivel its two children,
the left and the right leg. The leg is composed of the thigh on which is rooted the
calf.
The upper torso is rooted at the chest, which has has three children: the head,
the right hand and the left hand. The head is composed of the neck upon which
rotates the face. The hands can rotate beside the chest, and can be firther
decomposed into upper and lower arms. With this motion in mind, we can define
a hierarchy for the android as shown in Fig.5.35.
By this definition, if we move the Root component, the entire android (with
all its children) will be transformed. Swiveling the arms or legs will rotate the
entire a d l e g , but rotating only the lower arm or leg will have no affect on its
parents. Let us implement the Android model as a hierarchical model. The model
has been defined in a VRML file. Each component of the model has been
defined with a specific index in the coordinates data array, as shown below.

#define ROOT 20
Wefine CHEST 19
#define NECK 18
#define FACE 17
#define LEFTUPPERARM 16
#define LEFTLOWERARM 15
#define LPALM 14
#define LFINGERS 13
#define RIGHTUPPERARM 12
#define RIGHTLOWERARM 11
#define RPALM 10
#define RFINGERS 9
#define PELVIS 8
#define LEFllHlGH 7
#define LEFTCALF 6
#define LHEEL 5
#define LTOES 4
#define RlGHllHlGH 3
#define RIGHTCALF 2
#define RHEEL 1
#define RTOES 0

We abstract just the drawing routine for any part identified by the above-
defined component ID in the following function:

void drawWComponent1intcomponent) {
int i;
-
for li O;i~noofpoly[componentl"3;i -i + 3) {glBeginlGL-TRIANGLES);
gNertex3fvl&lcoords[component][3"indices[componentl[ill~~;
glVertex3fvl&lcoords[componentl[3*indicescomponentl+ 111));
glVertex3fvl&lcoords[componentl[3*indices[componentl+ 211));
glEnd0;)
1
Let us start defining the actual components. For this model, we do not define
an array of transforms; each component has only specific transforms that can be
applied to it. The right leg is composed of the calf as its leaf node (we do not
maintain the toes and heel as further nodes in this hierarchy). The calf can only
rotate about the x-axis (swing back and forth). Since the calf is not defined with
its base at the origin, we do need to translate it to the origin before rotating it (and
translate it back when we are done). This will ensure the correct pivot point
being used for rotation. The component Rightcalf is defined as follows:

void draw-RightCalflGLfloat angx) {

11 save the current transformation state
glPushMatrix0;
11 translate the component to the origin before rotation: piiot point
glTranslatef1-.14,-I .59,0.1);
glRotateflangx, 1.,0.,0.);
I1 translate the component back to the original posiion
glTranslatef(.l4,1.59,-0.1 I;
11 draw the actual geometry in this transformed space
draw ComponentlRIGHTCALF);
draw-~om~onentl~~~~~);
draw-~om~onent(~~~~~);
11 reset t?;e transformation stack
glPopMatrix0;)

The thigh can rotate about both the x- and z- axis for a forward and sideways
swing respectively. The function to draw this component is defined below. In
addition to its own transforms, we also pass in the rotation information for its
child node.

void draw RightThighlGLfloatangx,GLfloat angz,GLfloat calfangl {

g~~~~h~atrix~;
glTranslatef(-.2,-.7,0.1;
glRotatef(angx, 1.,0.,0.1;
glRotateflangz, 0.,0.,1 .I;
glTranslatef1.2,.7,0.1;
11 Draw the component and all its children in this transformation space
draw RightCalflcalfang);
draw-CO~~O~~~~~RIGHTTHIGH);
g~~opMatrixl);
1
The top-level component, RightLeg, does not have any geometry associated
with it and is defined as follows:

void draw RightLeglGLfloat "rightlegang) {

draw-~ i & t ~ h i ~ h l r i ~ h t l e ~rightlegang[l],rightlegang[2]);
an~[~l,
1
J
The left leg and the arms are similarly defined with the appropriate hierarchy.
The head consists of the neck and the face. We define the face to be a child of
the neck. We allow it to swivel about the y-axis, causing it to revolve from side
to side. We do not allow the neck any freedom to transform.
128 5 7 HIERARCHICAL
MODEIING USINGTRANSFORMATIONS

void draw_facelGLfloat angyl {

glPushMatrix0;
glRotateflangy, 0.,1.,0.l;
draw ComponentlFACEl;
g~~opMattix~);
1
void draw_Neck(GLfloat headangl {
draw Facelheadangl;
dra~~om~onentl~~~~l;
1
void draw-Head(GLfloat headangl{
draw-Necklheadangl;
1
Notice that we do not define any limits to the amount one can swivel the
components. For realistic motion you should clamp the rotations so you don't
end up twisting the body in weird ways!
And finally the android itself can be defined as follows:

void draw-AndroidlGLfloat "T, GLfloat "R,

GLfloat headang,
GLfloat "leftamang, GLfloat "leftlegang,
GLfloat "rightamang, GLfloat "rightlegangl

glPushMatrix0;
11 transform the entire android
glTranslatef(Tt01, Tt11, Tt211;
glRotateflR[Ol,1.,0.,0.l;
glRotatefIRt11,0.,1 .,O.l;
glRotateflR[21,0.,0.,1 .I;
11 draw the components and their children
draw-ComponentlROOTl;
{
draw-CornponentKHESTI;
draw-RightAmlrightarmangl;
draw Lefttrrnlleftarmangl;
drawI~eadlheadan~l;
1
 draw-RightLeglrightlegangl;
1
glPopMatrix0;
1
In Exarnple5-I 4, we make Android march by continuously changing the rotation
on the arms and legs.

void Displaylvoidl
{
-- -
if (cycle%2 01 {
rightarmang[Ol- 0.3; leftarmang[Ol+ 0.3; -
-- -- -
rightarmang[21--0.4; leftarmang[21+ 0.4;
rightlegang[Ol- 0.6; leftlegang[Ol+ 0.6;
rightlegang[21+ 1.2; leftlegang[21- 1.2;
1 else {
--
rightarmang[Ol + 0.3; leftarmang[Ol- 0.3;
rightarmang[2] + 0.4; leftarmang[21- 0.4;
--
-- --
rightlegang[Ol + 0.6; leftlegang[Ol- 0.6;
rightlegang[21- I.2; leftlegang[21+ 1.2;
1
pos+ +;
if lpos >-I001 {
-
pos 0;
cycle+ +;
1
glClear IGL-COLOR-BUFFER-BIT);
glColor3f 11.0,1.0, 1.01;
glLoadldentill;
draw AndroidlT, R, headang, leftarmang, leftlegang, rightarmang, rightlogang);
g~~~~~T;~~;
glutSwapBuffers0;
glutPostRedisplayll;
1
You can experiment with different kinds of transformations on the Android
to verify that the motion is consistent with the model hierarchy we designed.
Ideally, we would like to move the Android based on some user input. We leave
this part of the code as an exercise for the reader. The entire code can be found
in Exarnple5-14/Example5-14.cpp. You will have to compile and link this
program with the provided files: vrrnl.cpp and vrrn1.h.
Summary
In this chapter, we have learned the basics of 3D modeling. The polygon shape
was used to define primitive as well as complex shapes. We have learned how
principles of transformations can be used to define hierarchical models. By
developing 3D models, defining a viewing position, and a projection, we can
define a 3D world inside our computer. The camera just needs to be clicked to
photograph the world and display it onto the screen. In the next chapter we shall
see how to render this 3D world to generate realistic images.
Chapter 6
Rendering: Shading
and Lighting

In the last chapter, we saw how to model objects and represent them as wire-
frame models. Wire frame models depict the outer hull of the object but do not
convey a realistic image of the model. The next step in our quest for visual
realism is to paint our models so we can render realistic images. This process is
called rendering.

In this chapter you will learn all about the components of rendering:

Hidden surface removal

4 The CG reflectance model
Surface materials (including texture mapping)
Lights
Shading algorithms
6.7 What is Rendering?
Rendering 3D images is a multi-step process. First, we need to identify which
component surfaces (usually polygons) of the model are viewable from the
current viewpoint. This process involves back-face culling as well as identifying
surfaces obstructed by surfaces in front of them. Once visible surfaces have been
identified, we can simply assign a color to them and paint them. We saw this
approach to some extent in Chapter 1.

Fig. 6.1: A sphere rendered with a) a single color,

b) each polygon rendered with a single color, c) shaded smoothly

In most cases, however, we do not want surfaces to be colored with just one
color. Surfaces should appear shaded based on the amount of light that they
receive, as is the case in real life. To be able to simulate this effect, we need to
define material properties of the surface: not only its color but also how it
responds to light. This process is called shading. We need to define light sources
to light up the scene, which enable us to view it-a process called lighting. Once
the shading and lighting of a scene has been established, shading algorithms are
then used to finally render the images.
Rendering is a complex process and can be very time consuming. For
example, a typical image in the film Toy Story took anywhere between an hour
and 72 hours to render on a workstation! Different shading algorithms use
different methods and tricks to simulate the behavior of light and surfaces.
Obviously, which algorithm you use is based on your precise needs. An
interactive game needs shaders whose output may not be very sophisticated but
can be rendered quickly, whereas a blockbuster movie production can afford to
use complex shaders that can take days to render images but produces
spectacular results. We shall employ some simple shading algorithms in this
book and shall introduce you to some more complex ones in Chapter 7.
Let us first start with a more in-depth discussion of hidden surface removal.
6.2 Hidden Surface Removal
In the previous chapter, we looked at wire frame representations of models. The
models we constructed were "see throughy'-we could see through the polygons.
We learned how to eliminate the back-facing polygonal surfaces of our model
using a technique called back-face culling. But what about those surfaces that are
obscured by other objects in front of them as shown in Fig..?

Fig. 6.2: A sphere obstructing a cone: back-face culling is on

In a wire frame mode, it probably is still okay to see through the first object.
But when we draw our object as a solid surface, we want to see the object that is
in front. Compile and run Example6-1/Example6_I. cpp.

Fig. 6.3: A cone a. in front of the sphere, b. behind the sphere

This example shows a solid cone and a sphere. The cone translates back and
forth along the z-axis. Surprise, surprise- even when the cone is behind the
sphere, we still see it being drawn in front! We want not only the back surfaces
of the sphere and cone to be hidden from view, but also the portion of the cone
being obstructed from view by the sphere. The process to remove these surfaces
is called hidden surface removal. Back-face culling is part of this removal
process.
Z-Buffering
One of the simplest techniques to accomplish hidden surface removal is known
as z-buffering or depth buffering. In this process, a buffer in memory called the
z-buffer is used to keep track of the surface that is closest to the eye for any given
pixel. The closest object finally gets drawn onto the screen. Note that for solid
closed objects, this process will automatically remove back-facing surfaces also.
OpenGL provides z-buffering by way of a depth buffer. To use it, you need to
first enable depth buffering. You can do this by initializing the display mode to
use the depth buffer

glutlnitDisplayMode (GLUT-DEPTH I .... 1;

-and then enabling the depth-buffering test

Before drawing the scene, you need to clear the depth buffer and then draw
the objects in the scene. OpengGL will perform all the calculations to store the
closest surface information in its depth buffer and draw the scene accordingly.
Shown below is a snippet of code to draw a sphere and a cone. The cone is
animated to periodically move in front of and then behind ahead the sphere.
Depth buffering ensures that obstructed parts do not get drawn onto the screen.
The entire code can be found under Example6-I/Example6_I.cpp(you will need
to remove the commented lines to enable the depth test)

glClear(GL-COLOR-BUFFER-BIT I GL-DEPTH-BUFFER-BIT);
-
Tz Tz + Tinc;
-
if (Tz > 51 Tinc -0.5;
-
if (Tz < -5) Tinc 0.5;
glLoadldentity0;
gl~L00kAt(O.,O.,l0,0,0,-100,0.,1.,0.);
glRotatef(-90,1,0,01;
{
glColor3f(l.,O.,O.);
glutSolidSphere(2., 8,81;

When you make a call to glColor, all surfaces drawn after the call will use
the same color. If you see how the cone and sphere are rendered, you will be very
 6 RENDERING'
CHAPTER SHADING
AND LIGHTING 135

disappointed. They look so 2D! We would like our surfaces to be shaded to give
a sense of depth.
Surface removal is but one factor in generating a realistic image. Our next
task is to shade the visible surfaces so they appear 3D based on the lights
illuminating them, and their surface material properties. Since we want to
simulate the real world, CG simulation of lighting and shading follows a close
analogy to how lights and surfaces behave in the real world, and how humans
actually see. Let us see how light works in the real world and then draw upon
this analogy in our discussions of CG based lights and materials.

6.3 Light: Reflectance Model

Imagine entering a dark room with no lights in it. If the dark room is perfectly
insulated from any kind of light, you will not be able to see anything in the room
even though there are objects in it.
Imagine now that we switch on a bulb in the room. Immediately, we shall

Fig. 6.4: Light rays entering the eye enable sight

start seeing the objects in the room. How brightly we see them will depend on
the wattage and color of the bulb as well as on the material that the objects are
made of. The same object may appear reddish under a red light or eerily blue
under a blue light. We are able to see because light bounces off (reflects) from
objects and eventually reaches our eyes. Once light reaches our eyes, signals are
sent to our brain, and our brain deciphers the information in order to detect the
appearance and location of the objects we are seeing. The light sources directly
or indirectly define the incident light rays. .The surface properties of the objects
in the room, also called the surface material, determine how the incoming light
is reflected. Let us explore the way light is reflected in further detail.

Law of Reflection
Light is usually modeled as light rays (which can be thought of as vectors!).
Many rays traveling together are referred to as a beam of light. Light rays behave
in a very predictable manner. If a ray of light could be observed approaching and
reflecting off a flat mirror, then the behavior of the light as it reflects would
follow the the law of reflecti0n.A~shown in Fig.6.5, a ray of light, I, approaches

Fig.6.5: Law of reflection

the mirror and is called the incident ray. The ray of light, R, that leaves the
mirror, is known as the reflected ray. At the point of incidence where the ray
strikes the mirror, we can define a normal, N, to the surface of the mirror. The
normal is perpendicular to the surface at the point of incidence and divides the
angle between the incident ray and the reflected ray into two equal angles. The
angle between the incident ray and the normal is known as the angle of
incidence. The angle between the reflected ray and the normal is known as the
angle of reflection. The law of reflection states that when a ray of light reflects
off a surface, the angle of incidence is equal to the angle of reflection.
Whether the surface being observed is microscopically rough or smooth has
a tremendous impact upon the subsequent reflection of a beam of light. Fig. 6.6
below depicts two beams of light, one incident upon a rough surface and the
other on a smooth surface.

Fig.6.6: Reflection off smooth and rough surfaces

For both cases, each individual ray follows the law of reflection. For smooth
materials, such as mirrors or a calm body of water, all the rays reflect and remain
concentrated in the same bundle upon leaving the surface, causing the surface to
have a shiny look. This type of reflection is known as specular reflection.
For rough materials, such as clothing or an asphalt roadway, the roughness of
the material means that each individual ray meets a surface with a different
orientation. The normal line at the point of incidence is different for different
rays. Subsequently, when the individual rays reflect according to the law of
CHAPTER
6 RENDERING
SHADING
AND LIGHTING 137

reflection, they scatter in different directions. The result is that the rays of light
incident to the surface are difhsed upon reflection. This type of reflection is
called diffuse reflection. In practice, both types of reflection can be observed on
the same surface.
When light strikes an object, not only does a portion of it gets reflected, as
discussed above, but it can also be absorbed or transmitted through. Objects have
a tendency to selectively absorb, reflect, or transmit light of certain colors. So,
one object might reflect green light while absorbing all other frequencies of
visible light, causing us to see it as green. Another object might selectively
transmit blue light while absorbing all other frequencies of visible light, causing
it to look blue and translucent.
Light being bounced around tends to decay over time and distance due to
atmospheric absorption. This is why far-away objects appear dimmer than
objects that are closer to us. This phenomenon is referred to as attenuation.
The color and intensity of the reflected light that eventually reaches the eye
is the color that the surface is perceived to have.

6.4 CG: Reflectance Model

In CG we try to emulate the above-defined reflectance model to approximate
the way lights and surface materials. Mathematical equations that approximate
this behavior govern the final color of each pixel in the image. The most popular
CG reflectance model is the one proposed by Phong.

Phong Reflectance Model

The Phong reflectance model is also the model employed by OpenGL. In this
reflectance model, we break reflectance into four components:
1. Ambient Reflectance: This is a hack introduced in CG. It is used to
avoid CG scenes from going completely black. Ambient light has no
direction; it impinges equally on all surfaces from all directions and is
reflected by a constant multiple, resulting in flat-looking surfaces.
2. Diffuse Reflectance: Surfaces with diffuse reflectance scatter light
equally in all directions, but the intensity of the reflection varies based
on the angle of incidence.
3. Specular Reflectance: Specular reflections can be observed on any
shiny surface, causing a dull grey or white highlight on the surface.
4. Emission: The emission component of a surface is again a hack. It is
used to simulate luminous objects that seem to glow from their own
light, such as a light bulb.

Within our CG scene, we define surface materials with reflection coefficients

for each of the four components. The reflection coefficients are defined as RGB
triplets, and have a value between 0 and 1 - 1 being maximum reflection. Many
people refer to these reflectance coefficients as the color of the material, since it
will be the color of the light it will reflcct. This material is then assigned to
models in the scene.
Incoming
light

Specular
reflection

I Surface
I
Fig.6.7: Four components of Phong reflectance model

CG lights are defined to illuminate the scene. These lights are again modeled
with RGB colors for the ambient, diffuse, and specular components. The red,
green and blue components have a value between 0 and 1. Some software
packages further define an intensity to the light: the intensity is simply a scalc
factor for the color of the light: the actual color of the light is the product of its
color and its intensity. The OpenGL modcl does not use intensity for light sources.
When incoming light strikes an object, each of these four components is
calculated to detcrminc its individual contribution to the reflcctcd light. All four
components then added together to attain the final color of the outgoing light
Let us look into each of these components and how they are calculated.

Ambient Reflectance
Ambient reflectance is actually just an approximation to the fact that there is
never a perfect dark room in this world. Some amount of light always seeps in,
enabling vision. This kind of illumination enables us to see faint traces of objects
with flat shades of surface colors and is refcrred to as ambient lighting.
When light with an ambient component (ambient light) strikes a surface with
ambient reflectance, it is scattcrcd equally in all directions with a constant
multiple, causing the surface to look flat as shown in Fig.6.9.
CI-~AP'IEIH6 RENDERING.
SHADING
AND 1-IGHTING 139

Fig.6.8: Ambient Reflectance

Mathematically, if the incident light has a an ambient color of (Ri,Gi, Bi) and
the object it hits has a material with ambient reflectance of (Ra, Ga, Ba) then the
reflected light due to ambient reflection will have a color defined as
(Rra, Gra, Bra) = (Ri*Ra, Gi*Ga, Bi*Ba).

Fig.6.9: Ambient lighting on a sphere

Let us see how to render a sphere using ambient lighting. In openGL, we can
specify global lighting. Global lighting is not associated with any light source,
but is prevalent uniformly in the 3D world being defined. The command to set
up a global ambient light is

where lmodel-ambient is a vector defining the ambient RGBA intensity. A is the

alpha component of the color. We will not be using it in this chapter, so always
set this value to 1. By default, the value of the ambient color is

In Example6-2, we use global ambience to light up a sphere. The init function defines
the lighting model as ambient and sets the ambient color to a dull white, as shown below.
glLightModelfvlGL-LIGHT-MODEL-AMBIENT, light-ambient);

We need to enable lighting in order for it to affect the rendering as shown below

We also need to define a material for the sphere. To set material properties of an
object, we use the OpenGL command

void glMaterialfv(GLenumface, GLenum pname, TYPE" paraml;

This command specifies a current material property for use in lighting calcu-
lations.
face can be GL-FRONT, GL-BACK, or GL-FRONT-AND-BACK to
indicate which face of the object the material should be applied to.
pname specifies the particular material property being (such as
GL-AMBIENT for ambient reflectance)
param is the desired value(s) for the property specified.
In our example, we set the ambient reflectance of the material such that it
reflects all the red component of incident light and absorbs all of green and blue
by defining the material as

-
GLfloat mat-ambient0 { 1, 0.0,0.0, 1.0);
glMaterialfviGL-FRONT, GL-AMBIENT, mat-ambient);

Since it reflects only the red component of light, the material will look red.
Indeed, you can see the analogy between the color of the material and its
reflectance coefficients. Surfaces are rendered using the most recently defined
material setting. Since we will only be setting one material for this example, we
can set it in the init() finction itself. The display code is as follows:

void Display(void)
{
glClear (GL-COLOR-BUFFER-BIT I GL-DEPTH-BUFFER-BIT 1;
I1 material as already been set in the init function
glutSolidSphere(2., 40,321;
glhsh(l;
1
Remember to always turn on depth buffering when lighting 3D scenes. This
will ensure you are always viewing the surfaces that are closest to the camera!
Run the program. Notice how the sphere appears as a dull flat colored red ball.
The sphere looks flat because it only has ambient reflection defined. The
entire code can be found under Example6-2/Example6-2.cpp.
c l - 1TER
~ ~6 RENDERING.
SHADING
AND LIGFITING 141

Diffuse Reflectance
Surfaces with diffuse reflectance scatter light equally in all directions. The
amount of light reflected is directly proportional to the angle of incidence of the
incoming beam of light. Diffuse reflectance can be observed on dull objects such
as cloth or paper.

I Surface
I
Fig.6.10: Diffuse reflectance
The exact math to calculate diffuse reflection was proposed by Lambert and
so this reflectance is often referred to as Lambert reflectance. If we assume the
light to have a diffuse color of (Rid, Gid, Bid) and the surface to have a diffuse
reflection coefficient of (Rd, Gd, Bd). then the diffuse color component of the
reflected light can be calculated as

where 6 is the angle between the incident ray (direction of light source) and the
normal to the surface at the point of incidence. So if 8 is 0 (the light hits the

Fig.6.11: 8 used in the diffuse reflection calculation

surface straight on) then the diffuse reflection is the brightest, but light incident
on the surface at more than 90 degrees (cos(>90) <= 0) causes no light to be
reflected, and the corresponding area appears black. The diffuse color
component of a light essentially models the color of the light as we perceive it.

incident light

Fig.6.12: Spherical surface with diffuse reflectance

Let us apply ambient and diffuse reflectance to our famous snowman from
Chapter 5. In Example6-3, we define all parts of Snowy to have the same
ambient reflectance. The ambient color is slightly blue and is defined in the top-
level function, draw-Snowman as

-
GLfloat mat-ambient0 { 0.1,0.1, 0.2, 1.0 );
glMaterialfv(GL-FRONT, GL-AMBIENT, mat-ambient);

The diffuse components are different for each part. The snowballs have a whitish
diffuse component:

GLfloat snow-diffuse0 - {0.8,0.8,0.8,1.);

Snowy's eyes have a black diffuse reflectance his carrot nose is orange and his
stick hands are brown

GLfloat eye-diffuse0
GLfloat nose-diffuse0
--{0.0,0.0,0.0,1.);
{0.9,0.5,0.0,1.);
GLfloat hand-diffuse0 - {0.5,0.3,0.1,1.);
Surfaces are rendered using the most recently defined material setting.
Hence, we assign individual materials to Snowy's parts in their local drawing
routine. For example, the draw-Bottom function is now defined as
void draw-BottomlGLfloat "botXforms){
glMaterialfv(GL-FRONT, GL-DIFFUSE, snow-diisel;

and the draw-Nose function is redefined as

void draw-Nose(GLfloat *noseXformsl{

glMaterialfvlGL-FRONT, GL-DIFFUSE, nose-diffuse); etc.

We now need to define a light with a specific direction to see the effect of
difise reflectance. The OpenGL command to specify a light source and its
properties is either glLightf for nonvector proprties or glLightfv for vector
properties:

void glLightflGLenum light, GLenum pname, TYPE param);

void glLightfv(GLenum light, GLenum pname, TYPE "param);

The command(s) take three arguments.

light identifies the light source and can have one of 8 values: GL-LIGHTO,
GL-LIGHT1, ..., GL-LIGHT7.
pname identifies the characteristic of the light being defined. The
characteristics of lights include the component intensity, position and attenuation
of the light source, as well as certain spotlight settings.
param indicates the value that the characteristic is set to: it's a pointer to a
group of values if the vector version is used, or the value itself if the nonvector
version is used.
We define one light source, GL-LIGHT0 to light up our snowman. We define
this light source to have an ambient as well as diffuse color of dull white

GLfloat light-ambient0
-
- { 0.5,0.5,0.5,1.0 );
GLfloat light-diffuse0 { 0.8,0.8,0.8, 1.0 );
glLightfv(GL LIGHTO, GL-AMBIENT, light-ambient);
g l ~ i g h t-f v ( ~ ~ GL-DIFFUSE,
- ~ l ~ ~ ~ ~ ,light-diffuse);

We need to enable lighting as well as the light source chosen:

To set the direction of our light, a vector of four values (x, y, z, J

) is supplied with a
4
'
GL-POSITION parameter to the glLightfv hction. If the last value, W, is zero, the
corresponding light source is a directional one and the (x, J z) values describe its
direction. We define the direction of our light source to be coming along the z-axis:

GLfloat lightgosition0 - {O., 0.0, 1, 0.0);

144 6.4 CG. REFLECTANCE
MODEL

glLightfvlGL-LIGHTO, GL-POSITION, lightgositionl;

Note that the position and direction of lights are transformed by the
modelview matrix! Let us make our graphics more interesting by having the light
rotate about the sphere. To do this, we define the position of the light source in
the Display function (within the Model space), and constantly spin it around.

glPushMatrix 0; /I save current transform

glRotatef I spin, 0.0, 1.0,O.Ol; //rotation applied to light
gllightfv IGL LIGHTO, GL POSITION, lightgosition);
g l ~ o p ~ a t r iI/~ pop
i ; bacito saved transform
draw-SnowManl..l

The spin variable increments at every tick. The entire code can be found under
Example6-3, in files Example6-3 cpp, Snowman.cpp and Sn0wman.h.
Voila! You will see a snowman that is lit up by a rotating light source

Specular Reflectance
Specular reflection produces shiny highlights often seen on shiny objects like
glass or metal. Specular reflection, unlike ambient and diffise reflection depends
on both the direction of the incident light as well as the direction that the surface
is being viewed from. The reflection is brightest when the viewing direction is
parallel to the reflected light.
The equations to define spccular reflectance have been proposed by many
mathematicians. Each formula simulates the specular reflectance slightly
differently. However, in general, specular reflection depends on three components:
I .surface orientation, N
2.direction of light source, I
3. viewing direction, V
Assuming a function specular that depends on the above three factors, the
specular reflection color can be defined as
RI, ,G, )B, -IRis*R<specular0, GkG<specularO, BkB<specularlll

viewpoint

I Surface I
Fig.6.13: Specular reflectance
 6 RENDERING
CHAPTER SHADING
AND LIGHTING 145

Phong proposed that this function can be approximated to be ~ o s ( a )where

~, a
is as shown in Fig.6.13, and n is the specular exponent that determines the size
of the highlight. Other equations have been proposed by Blinn and others to
model specular reflections as well. Each equation output shiny highlights, but the
softness of the highlight differs.

Fig.6.14: Specular reflection

The GL-SPECULAR parameter affects the color of the specular highlight.
Let us see how to implement a specular lighting scheme in openGL.
Typically, a real-world object such as a glass bottle has a specular highlight that
is whitish. Therefore, if you want to create a realistic effect, set the
GL-SPECULAR parameter of the light to a dull white. By default,
GL-SPECULAR is defined to be (1.0, 1.0, 1.0, 1.0) for GL-LIGHT0 and (0.0,
0.0, 0.0, 0.0) for any other light.
We can modify Example6-3, by adding a specular component to our
GL-LIGHT0 as follows:

-
GLfloat light-speculadl {0.8,0.8,0.8, 1.0);
glLightfvlGL-LIGHTO, GL-SPECULAR, light-specularl;

The material for all Snowy's components have a common specular reflective
component, defined as

-
GLfloat mat-speculadl {1.0,1.0,1.0,1.0);
glMateriahlGL-FRONT, GL-SPECULAR, mat-specularl;

OpenGL allows you to control the size and brightness of the highlight using
the parameter: GL-SHININESS. You can assign a number in the range of [0.0,
128.01 to GL-SHININESS-the higher this value, the smaller and brighter
(more focused) the highlight. We define our material to have a shininess defined
as

-
GLfloat low-shininess0 (20);
glMateriahlGL-FRONT, GL-SHININESS, low-shininess);
When you run the program, you will see a shiny snowman as shown in
Fig.6.15

Fig.6.15: Snowy with diffuse and specular reflection

Emission
The emissive color of a surface adds intensity to the object, but is unaffected by
any light sources.

Fig.6.16: Emissive component of reflectance

Since most real-world objects (except lights) don't emit light, you'll probably
use this feature mostly to simulate lamps and other light sources in a scene. In
Example6-3, we can add a red emissive color to Snowy's nose (he has a cold,
poor fellow) as follows:

-
GLfloat mat-emission0 {0.5,0.0,0.0,0.0);
void draw-NoselGLfloat *noseXforms){
glMaterialfvlGL FRONT, GL-DIFFUSE, nose-diffuse);
g l ~ a t e ~ i-a l f v l ~GL-EMISSION,
~ - ~ ~ ~ ~ ~mat-ernissionl;
,

When you run the program, you will see a glow in the dark nose. Scenes can be
pretty boring with just one kind of light illuminating them. In real, we see lights
of all kinds--distant lights such as the sun, spotlights such as light emanating
from a light bulb or those used in theater. All these lights help to add to the three-
dimensionality of our world. Let us explore what more OpenGL can provide to
us in terms of lights.
 6' RENDERING'
CHAPTER SHADINGAND LIGHTING 147

OpenGL Lights
OpenGL can assign many more attributes to lights than just their color. Lights
can be defined with a position, a direction, and even a cone of illumination etc.
The most commonly used types of lights in OpenGL are
Distant (directional) light
Point light
Spotlight

Distant Light
The best example of a distant light is the sun. As viewed from any one position
on the earth, its rays are essentially parallel, and the position makes no practical
difference on the intensity of the light. In Fig.6.17 a distant light is shown light-
ing up a sphere. The distant light is represented by directed lines to indicate
which direction the light is coming from. Notice how the side of the sphere
facing away from the light is completely black.

Fig.6.17: Distant Light

The light rays from a distant light flow uniformly in space from one
direction. As a result, surfaces with the same orientation receive the same
amount of light independent of location. Surfaces with different orientations are
illuminated differently. Those facing toward the light source appear brightest,
while those facing away are completely un-illuminated by this light source
within the scene. In OpenGL, distant light sources are modeled with a color and
a direction. We saw an example of how to use a distant light when we were
illuminating our snowman in Example6-3.

Point Light
A bulb is a good example of a point light. A point light source distributes light
through space from a single point. It distributes beams of light evenly in all
directions. The intensity of the light is usually defined to fall off with the square
of the distance from the light to the surface. You could, however, define the
intensity to have no fall off at all. A point light needs a position to be completely
defined, and is often referred to as positional light.
 Fig.6.18: Point Light
In Example 6-4, we define a light source located at (-1,0,0) to light up a red
sphere (which is located at the origin). To define the position of the light, we set
the vector of (x,y,z, W)values for the GL-POSITION parameter. If W is nonzero,
the light is positional, and the (x, y, z) values specify the location of the light in
homogeneous object coordinates. This value is transformed by the model-view
matrix, so be careful where and how you define it!

I1 position of light
-
GLfloat light-position0 { -1.0, 1.0,0.0, 1.0 );
glLightfvlGL-LIGHTO, GL-POSITION, light~ositionl;
gluLookAtlO.,0.,5, O,O,-100,0.,1.,0.~;

The color of the light is defined in a manner similar to what we discussed earlier.
The entire code can be found in Example6-4/Example6-4.cpp.Try changing the
position of the light to see the results.

Spotlight
Theater lights or bulbs enclosed in a lamp shade are good examples of spotlights.
A spot light is like a point light, but its light rays are restricted to a well defined
cone. This kind of light is often used to direct the viewer's eyes to certain parts
of the scene. In OpenGL, you can use the GL-SPOT-CUTOFF parameter to
specify a light as a spotlight with the specified angle between the axis of the cone
and a ray along the edge of the cone. The angle of the cone at the apex is twice
this value, as shown in Fig.6.19. No light is emitted beyond the edges of the

Fig.6.19: A spotlight
 6 KRENDERING
CI~APTE SIIADINGAND LIGHTING 7 49

cone. The value for GL-SPOT-CUTOFF is restricted to being within the range
[0.0,90.0] (unless it has the special value 180.0). The following line sets
GL-LIGHT0 to be a spotlight with the cutoff parameter at 20 degrees:
glLightf(GL-LIGHTO, GL-SPOT-CUTOFF, 20.0);
You also need to specify a spotlight's position and direction, which
determines the axis of the cone of light:

-
GLfloat IightgosiiionU { -1.Or 1.Or 0.0, 1.0 );
-
GLfloat spot-direction1 { 0.0,0.0, -1.0 );
glLightfvIGL LIGHTO, GL-SPOT-DIRECTION, spot-direction);
g l L i g h t f-v ( ~ ~ - ~GL-POSITION,
l~~~~I lightgositionl;

Keep in mind that a spotlight's direction is transformed by the modelview

matrix just as though it were a normal vector. An example of a sphere lit by a
spot light can be found in Example6-5/Example6-5.cpp.

Fig.6.20: GL-SPOT-CUTOFF

Attenuation
For real-world lights, the intensity of light decreases as distance from the light
increases, a phenomenon known as attenuation. Since a directional light is
infinitely far away, it doesn't make sense to attenuate its intensity over distance,
so attenuation is disabled for a directional light. However, you might want to
attenuate the light from a positional (point or spot) light. To do this, you can
define an attenuation factor to your light. (OpenGL has three different factors for
attenuation. Here, we just look at the constant attenuation factor).

Refer to SHRE03 for more details on attenuation and how to set attenuation
factors in OpenGL.
150 6 4 CG. RFFLECTANCF
MODEL

Putting It All Together

The Phong model of CG lighting adds the above four reflectance components to
arrive at the color of the light that reflects from the surface. The point of
intersection between the reflected light and the image plane determines the pixel
being colored.
If there is more than one light source, then the contributions from each light
source are added together. The color of the pixel on the screen will be equal to
the final color of light that enters the eye/CG camera. The purpose of adding
lights to a scene is more than just to enable vision. It is to make the characters in

I Surface
I
Fig. 6.21: Adding all four components

the scene blend with their background, to enhance the mood of the scene, and to
make the scene eye catching. In Example6-6, we put Snowy against the
backdrop of the Alps. (Refer to Color Plate 2.) To make Snowy blend into the
scene, we need to identify and match the lighting of the background scene. The
scene seems to have a main light coming from the top right hand side. Although

Incoming
light

Fig. 6.22: More than one light

 this should be yellowish sunlight, the snow is causing it to appear white. A light
coming in from the top seems tp enhance the whiteness of the scene. A faint
bluish light will probably be reflecting off the snow to light up objects from
below.
To light up Snowy and a fake ground that he rests on, we set up three lights for
this scene:

--
GLfloat IightO-ambient0 { 0.5,0.5,0.5, 1.0 );
GLfloat IightO-diffuse0 { 0.8,0.8,0.8,1.0 );
GLfloat light0-speculafl-{ 0.2,0.2,0.2,1.0 );

GLfloat lightl-ambient0 - { 0.5,0.5,0.5,1.0 );

GLfloat lightl-diffuse0 -{ 0.3,0.3,0.5, 1.0 );

GLfloat light2-diffuse0 - { 1,1,1, 1.0 );

The final color of Snowy as seen in the color plate is the sum of the contributions
from ligh0, lightl, and light2. The code can be found under Example6-6.

6.5 The Normal Vectors

You may have noticed that both diffuse and specular reflection depend on the
normal vector to the surface. How and where is this normal vector being defined?
We saw in Chapter 5 that the normal vector to a polygon (or to a vertex on
the polygon) determines the orientation of the polygon. This normal vector can
be used to determine front- vs. back-facing polygons: polygons: with normal
vectors facing away from the viewpoint are back-facing and are culled from the
scene, as shown in Fig.6.23. The normal vector also determines the orientation
of the polygon or to a vertex on the polygon, relative to the light sources. This
normal vector is used in the lighting calculations we looked into earlier. In
OpenGL, we can define the normal vector by using the commands

+
Fig.6.23: Front- and back-facing polygons determined by the direction of normal vectors
 glNormal3l Gltloat nx, Glfloat ny, Glfloat nz);
void glNormal3fvlconst GLfloat *v);

This command sets the current normal to the value of the argument passed in.
Subsequent calls to glVertex"0 cause the specified vertices to be assigned the
current normal. For the glutSolidObject functions that we use, the normal vectors
are already identified within the routine.
In Example6-7, we read in the VRML model of the android by using our
ReadVRML function. Recall that this function also reads in the normal vectors
to the vertices of the polygons being defined.
In the Display function we assign normals to the vertices as shown:

glBeginlGL-TRIANGLES);
glNormal3fvl&lnormals[j][3*nindices[jl[ill)~;
glVertex3fvl&lcoords[jl[3*indices~lFll~~;
glNormal3fvl&lnormals[j][3*nindices[j][i+ I]]));
glVertex3fvl&lcoords~[3*indices[jl[i+ 111));
glNormal3fvl&lnormals~[3*nindices[j~ + 211));
glVertex3fvl&lcoords[jl[3*indices[jl[i + 211));
glEndl);

We also assign a shiny yellow material to the android. For each vertex,
OpenGL uses the assigned normal to determine how much light that particular
vertex receives and reflects, in order to generate the image. Run this program to
see the Android in all its metallic glory!
In the last section, we learned how to determine the color of a surface. For
every surface point, the four shading components are added together for each
light in the scene. This value is then clamped to the maximum intensity of 1 to
arrive at the final color of the point.
Performing this calculation for each surface point can be very tedious.
Consequently, shading models are used that limit the calculations to certain key
points on the surface. The surface is then colored by some average of these
values. For polygonal surfaces, the key points are the vertices of the defining
polygons.

6.6 Shading Models

Flat Shading Model:

In this technique, each surface is assumed to have one normal vector (usually the
average of its vertex normals) as shown in Fig.6.24. This normal vector is used
in the lighting calculations, and the resultant color is assigned to the entire
Ct-IAPTER6 RENDERINGSHADING
AND LIGHTING 153

surface. In Example6-7, we change the shading model of the Android to be by

calling

Notice how flat shading causes a sudden transition in color from polygon to
polygon. This is because every polygon is rendered using a normal vector that
changes abruptly across neighboring polygons. Flat shading is fast to render, and
is used to preview rendering effects or on objects that are very small.

( angle between the normal

and the incident light )
Incident Light

Viewing Direction (V)

Fig. 6.24: Flat shading: one normal per polygon

Gourad Shading
Gourad shading is used when smooth shading effects are needed. In this process,
each vertex defines a normal vector. This normal value is used to calculate the
color at every vertex of the surface.
The resulting colors are averaged into the interior of the surface to achieve a
smooth render. Since polygons share vertices (and hence the same normal value),

Camera

Fig. 6.24: Gourad shading: one normal per vertex

 the polygon edges blend together to produce a smooth look. Gourad shading pro-
duces smooth renders but takes longer to complete than flat shading. Even more
complicated rendering techniques exist, such as ray tracing, radiosity and volume
rendering. We shall look into some of these techniques in Chapter 9, when we
explore advanced concepts. All techniques a trade-off between computation time
and intricacy of the final render. In the last few sections, we saw how to assign
material colors to models and then render them with a desired shading scheme. As
detail becomes finer and more intricate, explicit modeling becomes less practical.
An alternative method often used is to superimpose a completely independent
image (either digitized or programmatically generated) over the surface of the
object. This method is called texture mapping.

6.7 Texture Mapping

Texture mapping can dramatically alter the surface characteristics of an object.
It adds vitality to models and can provide great visual cues for even simple
models. For example, if we map a woodgrain image to a model of a chair, the
chair will look like it is made out of fine wood grain. Texture mapping an image
of a brick wall onto a single polygon would give the impression that the polygon
had been modeled with many individual bricks, as shown in Fig.6.26.

Fig.6.26: Texture mapping a polygon

Texture mapping is a crucial element in today's games and graphic-oriented
programs. Without texture mapping, the models that are rendered would be far
from aesthetically pleasing. Because texture mapping is so usefkl, it is being
provided as a standard rendering technique both in graphics software interfaces
and in computer graphics hardware.

The Basics of 2 0 Texture Mapping

When mapping an image onto an object, the color of the object at each pixel is
modified by a corresponding color from the image. The image is called a texture
map and its individual elements (pixels) are called texels. The texture map
resides in its own texture coordinate space, often referred to as (s,t) space.
Typically (s,t) range from 0 to 1, defining the lower and upper bounds of the
image rectangle.
 6 RENDERING
CHAPTER SI-~AIING AND LIGHTING I55

The simplest version of texture mapping can be accomplished as described.

The corner points of the surface element (in world coordinates) is mapped onto
the texture coordinate space using a predetermined mapping scheme. The four
points in the (s,t) space defines a quadrilateral. The color value for the surface
element is calculated by the weighted average of the texels that lie within the
quadrilateral. For polygonal surfaces, texture coordinates are calculated at the
vertices of the defining polygons. The texture coordinate values within the
polygon are linearly averaged across the vertices.

Fig.6.27: Texture mapping: from texels to pixels

The texture space is defined to be from 0 to 1. For texture coordinates outside

the range [0,1] you can have the texture data either clamp or repeat over (s,t) as
shown in Fig.6.28.

S
Fig.6.28: A texture image, repeated along t and clamped along s

After being mapped to a polygon or surface and transformed into screen

coordinates, the individual texels of a texture rarely correspond to individual pixels
of the final screen image. Depending on the transformations used and the texture
mapping applied, a single pixel on the screen can correspond to anything from a
tiny portion of a texel (magnification) to a large collection of texels (minification),
as shown in Fig.29. Filtering operations are used to determine which texel values
 texture screen pixels texture screen pixels
Fig.6.29: Magnification and Minification of texels

should be used and how they should be averaged or interpolated. There are a
number of generalizations to this basic texture-mapping scheme. The texture
image to be mapped need not be two-dimensional: the sampling and filtering
techniques may also be applied for both one- and three-dimensional images. In
fact 1D texture mapping is a specialized version of 2D mapping. The texture may
not be stored as an array but may be procedurally generated. Finally, the texture
may not represent color at all but may instead describe transparency or other
surface properties to be used in lighting or shading calculations.

Mapping Schemes
The question in (2D) texture mapping is how to map the two dimensional texture
image onto an object. In other words, for each polygonal vertex (or some other
surface facet) in an object, we encounter the question, "Where do I have to look
in the texture map to find its color?"

Basic Mapping Scheme

Typically, basic shapes are used to define the mapping from world space to
texture space. Depending on the mapping situation, we project the object's
coordinates onto the geometry of a basic shape such as a plane, a cube, or a
sphere. It's useful to transform the bounding geometry so that it's coordinates
range from zero to one and use this value as the (s,t)coordinates into the texture

Fig.6.30: Planar Texture mapping: mapping an image to a vase.

 6 RENDERING.
CHAPTER SHADING
AND L I GITING
~ 157

coordinate space. For example, for a map shape that is planar, we take the (x,y,z)
value from the object and throw away (project) one of the components, which
leaves us with its two-dimensional (planar) coordinates. We normalize the
coordinates by the maximum extents of the object to attain coordinate values
between 0 and 1. This value is then used as the (s,t) value into the texture map.
Figure 6.30 shows such a mapping. Let us see how to specify texture maps in our
openGL environment. First, we need to specify a texture. The OpenGL command
to specify a two-dimensional image as a texture is

The width and the height of the image must be a power of 2. We shall look
into the arguments of this command shortly. To set the magnification and
minification filters in openGL, you use the command

glTexParamaterilGL-TEXTURE-2D, GL-TEXTURE-{MAGIMIN} -FILTER, filter);

where the value forfilter is GL-LINEAR by default, which would define a
linearly interpolated filter. GL-NEAREST takes the closest pixel in the texture
image rather than using interpolation, and hence is very quick to render. As noted
earlier, texture coordinates lie in the 0-1 range. When the coordinates go out of
this range, we would like the texture to either be clamped at the endpoints or
repeated. You can specify the wrapping of the texture in both s and t
independently by

where wrap can be GL-CLAMP or GL-REPEAT.

Finally you specify the texture coordinates at the current vertex location by
making the call

right before you make the glVertex call. Note that all 2D texture commands have
a corresponding 1D call. You also need to define the texture mode to be used
while rendering. GL-MODULATE modulates the current lighting and color
information with the texture image. GL-DECAL uses only the texture color. You
can set the texture mode using the glTexEnvi function:

glTexEnvilGL-TEXTURE-ENV, GL-TEXTURE-MODE, mode);

Let us work through an example of applying a planar texture on three kinds

of objects: a vase, a sphere, and a cube. The vase is a VRML model that we
downloaded from the Internet.
 Assuming projection along the z-axis, the mathematics to calculate the pla-
nar texture coordinates is simple. Given any vertex point (x,y,z) and the object's
x and y extents (XMIN, XMAX) and (YMIN, YMAX); we can define a function
to map the world coordinates to texture coordinates as:

void MapCoordinatesIGLfloatx,GLfloat y, GLfloat z, int mapping)

{
glTexCoord2flIx-XMIN)I(XMAX-XMINl,(y-YMINIY- YMINII;
1
Now, before defining the polygon vertices, we make a call to this function to set
the corresponding texture coordinates. For example, we define the vertices of our
VRML object as follows:

MapCoordinates(coords[3*indices[i]], coords[3*indices[i] + I],coords[3*indices[il+ 21, mapping);

glNormal3fvl&lnormals[3*nindices[ill~~;
glVertex3fvl&lcoords[3*indices[ill~I;

In Example6-8, we define the planar texture for the three models. In the init()
function, we first generate and bind a new texture:

We then set the various wrap and filter parameters for this texture:

glTexParameterilGL-TEXTURE-2D,GL-TEXTUREEWRAPPS GL-CLAMP);
glTexParameterilGL TEXTURE 2D,GL TEXTURE WRAP TI GL CLAMP);
g l ~ e x ~ a r a m e t e r i l ~ ~ ~ G~ ~E TXETX~T ~U ~R ~ M2 ~A G
I > I L T EGL-NEAREST);
~
glTexParameterilGL-TEXTURE-2D, GL-TEXTURE-MI N-FILTER, GL-NEAREST)

Next, we actually read in the texture image-which is nothing more than a

BMP image file. The image size is a power of 2, namely, 5 12 by 5 12

bits - ReadBitmapl'checktexture.bmp", &info);

We define the loaded bitmap to be the current texture map.

Finally we enable texture mapping and the texture environment.

glEnablelGL TEXTURE 2Dl;

gITexEnvf1Gt-TEXTUIE-ENV,GL-TEXTURE-ENV-ODE, GL-MODULATE);
 The Display routine calls the drawing routines for the texture mapped
objects. The complete example can be found under Example6-8 in files
Example6-8.cpp, Shapetextures.cpp and Shapexh. You will need to compile the
program with the vrml.cpp, vrml.h, bmp.cpp and bmp.h files provided as well.
The texture image can be found under the Images directory.
Other bounding shapes can be used to make the mapping work more like a
shrink-wrap. Refer to [FOLE95] for more information on other mapping schemes.

Environment Mapping
If you look around your room you may realize that many objects reflect their
surroundings. A bottle of water, a mobile phone, a CD cover, a picture frame, etc.
are only a few examples of reflecting objects that could be found in any 3D
scene. To make the 3D world more realistic, objects should show reflections of
their environment. This reflection is often achieved by a texture-mapping
method called environment mapping.
The goal of environment mapping is to render an object as if it were reflec-
tive, so that the colors on its surface are those reflected from its surroundings. In
other words, if you were to look at a perfectly polished, perfectly reflective silver
object in a room, you would see the walls, floor, and other objects in the room
reflected off the object. (A classic example of using environment mapping is the
evil, morphing Cyborg in the film Terminator 2.) The objects whose reflections
you see depend on the position of your eye and on the position and surface angles
of the silver object. Of course, objects are not usually completely reflective, so
the color of the reflection is modulated with the object's actual color for a
realistic look.
True environment mapping can be achieved using ray tracing, a technique we
shall learn about in Chapter 7. Ray tracing is a very expensive option and is
usually not necessary for most cases. More often, certain tricks are used to

Right

u Bottom

Fig.6.31: Six textures for cube mapping

 Fig.6.32: Cube Mapping
achieve reflective effects. A method often employed is called cube environment
mapping. The idea behind cube environment mapping is very simple. From an
object with a reflective surface you first generate six images of the environment
in each direction (front, back, up, down, left, right). That is, imagine placing a
camera at the center of the object and taking the size photographs of the world
around it in the directions specified. Based on the normal vector of the vertex in
consideration, the appropriate image is selected. This image is then projected
using planar mapping onto the object as shown in Fig.6.32.
The resultant object looks like it is reflecting its environment!
Mathematically, the cube face is identified by using the normal vectors
(nx,ny,nz) on the surface of the object. The greatest component is used to identify
where the surface should be "looking"; and hence the cube face and the texture
image to be used. The other two coordinates are used to select the texel from the
texture by a simple 3D to 2D projection. If, say, ny was the highest value
component then we divide the other components by ny (nx/ny,nz/ny). These
coordinates are normalized to give the (s,t)values into the texture map. The code
to find the face and bind the appropriate texture image is shown below:

int BindTexturelGLfloatnx, GLfloat ny, GLfloat nz1 {

-
GLfloat normu {nx,ny,m);
normalizelnorml;
- -
if I (fabslnorm[ll1 > fabs(norm[2111 && Ifabs(norm[l]b fabs(norm[O]l11{
if (norm[ll > -0.1 {
11 top only
glBindTexturelGL-TEXTURE-2D, t[011;
return TOPFACE;
) else{
11 bottom only
glBindTexturelGL-TEXTURE-2D, t[ll1;
return BOTTOMFACE;
 I
) else if (fabs(norm[21I> fabs(norm[O]II{
-
if (norm[2] > 0.1 {
glBindTexture(G1 TEXTURE_2D, t[2]1;
return FRONTFACE;
) else {
glBindTexture(GL_TDCTURE_2D, t[3]I;
return BACKFACE;
1
) else
{
-
if (norm[Ol> 0.1 {
_
glBindTexturelGL TEXTURE_2D, t[41);
return RTFACE;
1
else{
_
glBindTexturelGL TEXTURE2Dr t[5]);
return LTFACE;
1
1
return -1;
1
The code to use planar mapping to texture-map the coordinates is as follows:

void MapCoordinateslGLfloat nx,GLfloat ny, GLfloat nz, int face)

{
if (face -- TOPFACEI
glTexCoord2fllnxfny+ 1I12.,11 -nz(nyI12.1;
else if (face --BOTTOMFACE)
glTexCoord2flll -(nxlnyIIl2.,(1 -Inzlny)Il2.I;
else if lface --FRONTFACE]
glTexCoord2f(lnxlnz+IIR.,(nylnz+ 1112.);
else if (face --BACKFACE]
glTexCoord2flll+ nxlnzll2.,ll -nylnzIl2.);
else if lface --
RTFACEI
glTexCoord2f(ll -Inz~nxIIl2.,(nylnx+1112.);
else
glTexCoord2fl(-(ndnx)+ 1)12.,(-(nylnxl+ ll12.I;
1
Before making a call to begin a polygon, we first find the face of the cube
that this polygon will index and then call Mapcoordinates to generate the correct
(s,t) value into the texture. In Exarnple6-9, we read in the six texture images
 shown in Color Plate 3 as the six faces of the cube, and use them to environment
map the vase, sphere and cube. The final code can be found under Example6-9.
You will also need vrml.cpp, vrml.h, bmp.cpp, bmp.h and utils.h to compile and
run this program. The textures can be found under the Images folder under the
installed directory for the sample code. The rendered objects are shown in Color
Plate 4.
All the examples above use the vertednormal data in object space to map
into the texture. This means that if our object moves in the 3D space, the texture
will be 'stuck' onto it. This may not be desirable in the case of environment
mapping. As the object moves in the world, you want the reflections to change
accordingly. In order for the map to move, you need to pass in the vertednormal
data modified by the current transformation space. Use the OpenGL command

to get the current model view matrix and apply the necessary matrix operations
on your vertexlnormal data. Interested readers should try using this setup to map
the cube environment onto our marching android from Chapter 5.

6.8 Verfex Shaders

No discussion on rendering would be complete without talking about vertex

shaders. The concept of shaders is nothing new to the graphics industry, Pixar
has used it for years in their phenomenal films such as Toy Story and The
Incredibles. However till now, vertex shading effects were so computationally
complex that they could only be processed offline using server farms. With the
current generation of semiconductor technology, developers can program real-
time Vertex Shaders to breathe life and personality into characters and
environments, such as fog that dips into a valley and curls over a hill; or true-to-
life facial animation such as dimples or wrinkles that appear when a character
smiles.
A vertex shader is a graphics processing function used to add special effects
to objects in a 3D environment by performing mathematical operations on the
object's vertex data. Vertex data refers to data of the points along the surface, at
some specified precision (and not just at the vertices of the defining polygons!)
Each vertex data is defined by many variables. For instance, a vertex is always
defined by its location in a 3D environment using the x-, y-, and z- coordinates.
Vertices may also be defined by their normal vector, color, texture, and lighting
characteristics. Vertex Shaders don't actually change the type of data; they
simply change the values of the data, so that a vertex emerges with a different
color, different textures, or a different position in space, and is rendered as such
on the screen.
A vertex shader can be thought to be a magic box. Vertex data goes in and
 6'RENDERING:
CHAPTER SHADING
AND LIGHTING

Vertex Data ( i n S Vertex Data (outL render on

Position Position screen
Normal Normal
Color Color

Fig.6.33: Vertex shader

modified vertex data comes out. What the box does depends on how the
programmer develops the shader.
With the release of OpenGL1.4, the OpenGL group has released the OpenGL
Shading Language. The OpenGL Shading Language has been designed to enable
programmers to express vertex shaders in a high level language.
Independently compiled programs that are written in this language are called
shaders. A program is a set of shaders that are compiled and linked together.
OpenGL provides entry points to manipulate and communicate with these
shaders and apply them to objects within the CG world being defined for
stunning looking renders.
Vertex shaders is a vast topic, and deserves an entire book to do it full justice.
Refer to [ROST04] for more information on this topic. We are confident that
with the concept you have learned in this book, vertex shaders will not be
difficult for you to grasp.

Summary
In this chapter, we have studied the important techniques used in rendering. Our
quest for visual realism has taken us from hidden surface removal to lighting and
materials to texture mapping of surfaces. These techniques approximate the
physics of sight to generate photorealistic images. We shall look into some more
advanced rendering algorithms in the next chapter.
Chapter 7
Advanced Techniques

Graphics scenes can contain different kinds of objects: trees, flowers, glass, fire,
water, etc. So it is not too surprising that numerous techniques can be applied to
depict the specific characteristics of each object. In the last few chapters, we
learned about the basic principles of modeling and rendering. In this chapter, we
shall look into some advanced techniques popular in the CG world.
In Chapter 5, we dealt with linear components (lines and polygons) to depict
the hull of models. For complex models, polygonal surfaces can be tedious to
build for and may also occupy large amounts of storage space. Curves and sur-
faces are often used instead in order to achieve smoother representations with
less storage space. In this chapter, we look into several kinds of curves and
surfaces used in CG. Subdivision surfaces is another technique that has become
popular to achieve a high degree of local refinement.
In Chapter 6, we saw some basic rendering techniques. Although we went
from wire-frame to real shaded images, we are still not close to the photo-
realistic images that we set out to produce. Our images don't have shadows or
transparency, for example. In this chapter, we shall also talk about an advanced
algorithm called ray tracing to achieve such effects. Radiosity is another
rendering technique that we shall discuss briefly.
In this chapter, you will learn
w Advanced modeling
Curves-Bezier, B-splines, and Nurbs
Nurbs surfaces
Other techniques
Advanced rendering algorithms
Ray tracing
Radiosity
7.7 Advanced Modeling
Numerous techniques are used to model objects in CG. Which technique is used
depends on the object being modeled, as well as the final outcome desired.
Polygonal models are quick and easy to display but do not necessarily depict a
smooth surface. They are used widely in the gaming industry for speed of
display. Other techniques can depict smooth hulls but are slower to display. One
such technique is the use of parametric curves and surfaces.

Parametric Curves
Curves are used in CG for all kinds of purposes-fonts, animation paths and
modeling, to name a few. It is instructive to look into the mathematical equations
used to represent curves to understand how they are used in CG.

Parametric Equations
If you recall high school algebra, the implicit function for a curve in 2D space is
of the type:
fky) =0
For example a circle with radius r centered at 0 has a defining equation
x2+yL2=o

Fig.7.1: Curve segments meeting at a join point

All points (x,y)that satisfy the above equation are along the path of the circle.
A line can be thought of as a specialized curve-it's a linear curve! A line
with two endpoints, one at ( x l , y l ) and the other at (x2,y2) has the implicit
equation
y-m*(x-x1)-yl= 0
where m = ('y2-yl)l(x2-xI)and is also referred to as the slope of the line.
All points, (x,y) that satisfy the above equation are positioned on the line
being defined. The same curvelline can also be represented using a parametric
function. Parametric functions define each coordinate against a parameter of
choice. We saw in Chapter 1 , that our circle can be represented against the
parameter 8 as
x=rcos8, y=rsin8
As 0 vary's from 0 to 360, we evaluate the points along the circle.

Fig.7.2: varying 8 from 0 to 360

Our line can also be defined in a parametric form against a parameter t as

x = (x2 - xl)*t + x l
y = 0.'2 - yl)*t + y l
as t varies from 0 (defining the endpoint (x1,yl) to 1 (defining the
endpoint(x2,y2)) we evaluate the points along the line. It can be shown that in a
3D space, we can define a line against a parameter t as
x = axt + dx
y=a t+d
Y
z = azt +4
At t =0, we attain the endpoint of the line (dx,dydz) and at ~l we attain the
endpoint (ax+d, ay+dy az+dz). These two end points are enough to define the
entire line. In other words, they define the constraints of the line, and are also
called the control points of the line.
In Chapter 1, we saw how to use a set of lines (line segments) to approximate
the representation of a curved object-the circle. In CG, we say that the curve
(the circle in this case) is being approximated by linear segments. The segments
are called linear, because they are defined by an equation which is linear in
parameter t (t raised to the power of 1). In Chapter 5, we saw how a set of
polygons can approximate a curved surface such as a sphere-a linear
approximation to a curved surface.
Linear approximations to a curve (or a surface) usually requires a large
amount of vertex and normal data for reasonable closeness to the smooth
curve/surface. Defining a large number of vertices to approximate smooth curves
and surfaces is not only tedious but also error prone and difficult to manipulate.
Pierre Bezier was the first to develop a set of parametric cubic equations to
represent curves and surfaces using only a small set of control points. Many
more functions have been developed that produce better results at the cost of
efficiency. Keep in mind that these functions still only approximate the
surface/curve, but do so with less storage and offer easier manipulation than their
linear counterparts.
Cubic Curves
Curves can be approximated by a set of segments. We have seen how to use of
segments that are linear in nature. But the approximating segments need not be
linear. They can be quadratic (the defining parametric polynomials are t raised to
the power of 2), cubic (t raised to the power 3), or even higher orders. In practice,
cubic polynomials are the easiest to use and control and are very popular in CG.
If we refer to each segment of the curve as Q(t) = (x(t), y(t), z(t)), the cubic
polynomial equations defining the (x,y,z) points along the curve are of the type:
Cubic polynomials have four coefficients (a,b,c and d), and hence a curve
segment needs four constraints (control points) to define it.

The different kinds of curves (and surfaces) are classified by the coefficient
values of the polynomial equation. The coefficient values determine how the
curve interpolates the control points. Some of the commonly used curves are
Bezier, B-Spline, Hermite, etc. The two kinds of curves we shall study in this
book are Bezier curves and B-splines (specifically Nurbs).

x(t)
Fig.7.3: Curve segments meeting at a join point
Since a curve is approximated by more than one curve segment, it is
important to understand how the curve segments behave at the point at which
they meet which is called a join point. If two curve segments of a curve meet
together, the curve is said to have CO continuity. Usually, a curve should at least
be CO continuous to define a continuous curve! If in addition the tangent vectors
of the two curve segments (the first derivative, also called as velocity of the
curve) are equal at the join point, the curve is said to be CI continuous. If the
curvature of the two curve segments (second derivative, also called rate of
change or acceleration of the curve) are also equal at the join point, then the
curve is said to be C2 continuos as shown in Fig.7.3. C1 continuity is required
in curves to guarantee a smooth curve. C2 continuity guarantees a smooth
transition between segments.
Splines and, in particular, Nurbs have C1 and C2 continuity at their join
points and are often preferred over other types of curves.

Bezier Curves
The (cubic) Bezier curve segment has four control points: two control points, P1
and P4, define the endpoints of the curve, and the two control points, P2 and P3 Y
affect the shape of the curve by controlling the end tangent vectors.

Fig.7.4: Bezier curve defined by four control points

As shown in Fig.7.4, the starting and end tangents are determined by vectors
PIP2 and P3P4 We say that the Bezier curve interpolates (passes through) the
first and last control point and approximates the other two. Let us look into how
we can define and display Bezier curves using OpenGL.
The OpenGL evaluator functions allow you to use a polynomial equation to
produce vertices, normals, texture coordinates, and colors. These calculated
values are then passed on to the processing pipeline as if they had been directly
specified. The evaluator functions define a Bezier curve and are also the basis for
the NURBS (Non-Uniform Rational B-Spline) functions. The first step in using
an evaluator is to define the polynomial map by using the OpenGL command

In Example7-1,we define 4 control points for the Bezier segment, as

GLfloat ctrlPoints[41[3] -
{-6.10.10.1
-4.,6.,0.,
4.,-6.,0.,
 6.,0.,0.);
-
int numPoints 4;
The control points in the cubic map are defined by:

glMaplf(
GL-MAPI-VERTEX-3, I1 target, type of control data, 3 coordinates for each vertex
o., 11 lower t range
I., I1 upper t range
3, I1 stride
numPoints,&ctrlPoints[01[01~;

The parameters to this function are as follows:

target: The target defines the type of control points being used.
GL-MAP1-VERTEX-3 says that each control point has three floating-point
values representing x, y, and z. You can also use GL-MAP1-VERTEX-4 (for
x,y,z,w). There are other predefined values to determine the normal vectors and
texture coordinates for the curve as well.
trange: The lower and upper t range specifies how much of the curve you
wish to evaluate. t ranging from 0 to 1 defines the entire range of the curve.
stride: Stride represents the number of floats or doubles between the
beginning of one control point and the beginning of the next one in the data
structure referenced in ctrl points. This allows control points to be embedded in
arbitrary data structures. The only constraint is that the values for a particular
control point must occupy contiguous memory locations.
numpoints, ctrlPoints: Finally, we define the number of control points and the
array containing the control points.
We need to enable openGL to calculate the polynomial being mapped. We do
this by calling

The next step is to determine the domain values to be used to evaluate the
polynomial. OpenGL works with the concept of grids to specify evenly spaced
domain values for evaluation. The command

llmap a grid of 20 points along the t domain ranging from 0 to 1

glMapGridldl20.,0.,1.0~;

tells openGL to lay a grid of 20 evenly spaced points along the t domain. Finally,
we need to instruct OpenGL to evaluate the curve at the defined grid points. This
is accomplished by the command
 All the evaluated points are then smoothly connected using the primitive
specified (in this case, a line). You also need to specify the first and last integer
values for grid domain variable t. In this case, the curve will be evaluated and
plotted at t =(0,1/20,2/2- ... 1).
The more the number of points in our grid, the smoother will be the final
curve that is output. When you compile and run Example 7-I/Example7-l.cpp,
you will see the Bezier curve, defined by the control points specified.
Try changing the number of grid points to see how this affects the
smoothness of the curve. Note that this only changes how many points are used
to evaluate the polynomial equation-it doesn't change the equation being used.

Fig.7.5: Piecing together Bezier curve segments

Two Bezier curve segments can be constructed to have CO and Cl continu-
ity if we can ensure that they share a join point (P4) and the tangent vectors at
P4 are equal or Pg - P4 = P4 - Pj as shown in Fig.7.5. Try to define another
Bezier segment and connect it to the segment we defined in Example7-I. How
will you define the control points to achieve CO continuity? To achieve C1 and
C2 continuity?
One can approximate entire curves by assembling Bezier curve segments in
this manner. For more complex curves, the mathematics to ensure C 1 continuity
across the curve while adjusting control points gets tricky. A better solution is
provided by splines and by what we call Nurbs.

B- Splines
Spline curves originated from flexible strips used to create smooth curves in
traditional drafting applications. Much like Bezier curves, they are formed
mathematically from piecewise approximations of cubic polynomial functions.
B-Splines are one type of spline that is perhaps the most popular in computer
graphics applications. The control points for the entire B-spline curve are defined
in conjunction. They define C2 continuous curves, but the individual curve
segments need not pass through the defining control points. Adjacent segments
share control points, which is how the continuity conditions are imposed. For
this reason, when we discuss splines, we discuss the entire curve (consisting of
its curve segments) rather than its individual segments which must then be
stitched together. Cubic B-splines are defined by a series of (m=n+l) control
points: Po, P1...Pn
Each curve segment of the spline Qi 3 5 i,< n
is defined by the four control points Pi-3,Pi-2,Pi-1,Pi.
For example, in Fig.7.6, we show a spline with m=8 control points. The
individual segments of the curve are Q3, Q4, Q5, Q6 and Q7. Q3 is defined by
the four control points, PO-P3,Q4 by points PI-P4etc.
Conversely, every control point affects four segments. For example, in the

P2 P5 '6
I I

Control
Knot

I I
Q5 P4 P7
P3

Fig.7.6: A uniform nonrational B-spline

above figure, point P4 affects segments Q3, Q4, Q5, and Q6. Moving a control
point will affect these four segments but will not affect the entire curve. This is
a very useful property of B-splines that we shall look into more in the next
few chapters.
The join points between the segments are called knots. The knot between
segment i and i+l is represented as ki.The initial point of the first segment and
the endpoint of the last segment are also called knots, so there is a total of (n-I)
knots for the spline under consideration. When knots are uniformly placed, as
shown in Fig.7.6, the curve spline is called a uniform non-rational spline.
Unfortunately, it is difficult to define and control the splines since the segments
do not interpolate the control points.
Non-uniform non-rational B-splines define (n+5) knots. The knots need not
be uniformly spaced-and in fact are user defined. The advantage is that we can
force the curve to interpolate certain control points. Non-uniform B-splines uses
the notion of knot value sequences: a non-decreasing sequence of knot values
 that defines the placement of knots of the curve. For example, if we assumed the
curve above was a non-uniform non rational B spline, its knot sequence would
effectively be (0,1,2,3,4,5,6,7,8,9,10,11). (n+5 = 12 knots)
If successive knot values are equal in the sequence, it is called a multiple
knot. Multiple knots causes the curve to approximate the associated control point
more closely. In fact, three successive knot values forces the curve to actually
interpolate the control point, thereby making the shape of the curve easier to
define. Defining multiple knots does lead to a loss in continuity, but only at the
associated control point. If we modified our curve above to assume a different
knot sequence, the results would appear as shown in Fig.7.7.

Po.'-"LJ'-
Triple knot p7
Control Point
Knot
Fig.7.7:Non-uniform non-rational B-splines. Multiple knots, double knot
with knot sequence (0,1,2,3,4,4,5,6,7,8,9,10) causes C1 continuity,
whereas a triple knot with knot sequence ( 0 , 1 , 2 , ~ , ~ , ~ , ~ , ~ ,results
6,~,8,~)
in only CO continuity.

Nurbs
B-splines (or any other non-rational curve) can be defined in homogenous
coordinates by adding W(t)=l as the fourth element in the parametric equation;
Q(0 = (X(t), Y(0, Z(t), W(t)). (As usual, moving from homogenous coordinates to
3D space involves dividing by W(t)). This process is called rationalizing the curve.
The advantage of rational curves is that they are invariant under rotation, scaling,
translation and perspective transformations. This means we need to apply the
transformations only to the control points and then re-evaluate the curve equation
to generate the transformed curve. A non-uniform rational B-spline is also called a
Nurb and is heavily used in CG due to the properties described above.
In Example7-2, we draw a Nurbs curve using OpenGL. We first need to cre-
ate a new Nurbs renderer before we can define the actual curve. This is done by
using the command
We can set properties of this nurb as follows

gluNurbsProperty(pNurb, GLU-SAMPLING-TOLERANCE, 25.f);

gluNurbsPropertylpNurb, GLU-DISPLAY-MODE, GLU-LINE);

This property means that the resultant Nurbs curve will be drawn using a line
primitive. We define a set of eight control points (n= 7)

GLfloat ctrlPoints[81[31
{-5.,-5.,0.,
-
-5.,5.,0.,
0.,5.,0.,
O.,-5.,0.,
5.,-5.,0.,
8.,5.,0.,
12.,5.,0.,
12.,-5.,0.);

and knot value sequences of 12 (n+5)elements, with 1, 2, and 3 multiple knots,

respectively.

GLfloat Knotsl[l21
GLfloat Knots2[121
-- {0,1,2,3,4,5,6,7,8,9,10,11);
{0,1 ,2,3,4,4,5,6,7,8,9,l 0);
GLfloat Knots3[121 - {0,1,2,3,4,4,4,5,6,7,8,9);
The OpenGL command to define and display a Nurbs curve is

The parameters for this fimction are similar to the parameters we set for
glMap but includes support for defining the knot sequence.To define the curve,
we bracket it between gluBegnCurve and gluEndCurve.

gluBeginCurvelpNurb);
gluNurbsCunre(pNurb,
12, /Ithe number of knots
Knots1, I/the knot value sequence
3, //stride
&ctrlPoints[Ol[Ol, /Ithe control points
4, /Ithe order of the curve
GL-MAPI-VERTE-3); /I type of vertex data
gluEndCurve(pNurb);
 The order of the NURBS curve equals the degree of the parametric
equation+l; hence a cubic curve has an order of 4. The entire code can be found
in Example7-2/Example7-2.cpp. Try experimenting with different knot value
sequences to see what results you get. Interested readers are encouraged to refer
to DEB001 for more details on splines.

Parametric Bicubic Surfaces

Parametric bicubic surfaces are a generalization of the cubic curves. The surface
is partitioned into parametric surface segments or patches, which are stitched
together to form the entire surface.
The parametric equation for each patch is defined along two parameter
domains - s and t, defining a surface as opposed to a curve.

Fig.7.8: CO and C1 continuity at surface boundary.

Each patch is defined by blending the control points (4 *4 = 16 control points
are needed to define a patch). In fact, you can think of the control points defining
curves at specific domain intervals that are then meshed together to define a
surface.

'4 1

Fig.7.9: Control points for a patch

The continuity constraints on surfaces are the same as those for curves. CO
continuity implies that the two patches meet at the boundary curve. C1 continuity
implies that the two patches have a smooth transition at the boundary curve, etc.
The basis function used for Bezier curves can be extended along two
parameter domains to define Bezier surface patches. These surfaces are not very
 Control point polyhedra Control point polyhedra

Four sets of three control points

e
must be collinear

Boundary

Patches Patches

Fig.7.10: CO and C1 continuity at surface boundaries of 2 patches

popular due to the dificulty in stitching together continuous surfaces.
Nurbs surfaces can be defined as a logical extension of the Nurbs curve
equations we saw earlier. These surfaces have the same properties of Nurbs
curves-C2 continuity, their ability to introduce discontinuities through knot
placement and their invariance under transformations. As a result they are widely
used in depicting the outer hull of surfaces.
In Example7-3, we define a nurbs surface and periodically move the control
points to create a wave effect. We define four control points along the t domain,
for each of the seven slices along the s domain. The points define a rectangular
grid in the x-z plane within a boundary of (-12,12) along the x- and (-10,lO)
along the z-axes.

I1 s t lx,y,zl
GLfloat ctrlPoints[7][41[31 -
{{{-12rollo}r{-12ro15}l{-12rol-5}r{-12101-lo}}l 11s-0
{{-810110}1{-8r0r5}r{-8.r0.1-5.}1{-8.101-10}}1 11s-I
{{~4-~0-~10~}~{~4~0~5}~{~4r0~~5}~{~4~0~~10}}~/1~~2
{{O.,O.,l 0.~,{0,0,5}1{-0,0,-5}I{0101-1
0}}IIIs-3
{{4-~0~~10~)~{4~0r5}r{4r0~-5}~{4~0r~10}}J1~~4
{{8.,0.,10.},{8,0,5}r{8rOr-5}I{8101-10}}Jl~-5
{{12.,0.,1 0.}l{1210r5}r{1210r-5}l{12101-10}}ll~-6
1;
The knots along s and t are defined as

--
GLfloat KnotstJ81 {0.,0.,1.,1.,2.,2.,3.,3.};
GLfloat Knotss[lll {0,0,0,0,1.,2.,3.,4.,4.,4.,4.};

The knot sequence defines a surface that interpolates the first and last set of
control points along the s domain, as shown in the Fig.7.11.

Fig.7.11: The Nurbs surface

The surface is defined by the OpenGL call

This function has parameters similar to the curve but needs parameters against
two domains, s and t.

gluBeginSurFacelpNurb1;
gluNurbsSurface(pNurb,
11,Knotss,
E,Knotst,
4*3, llstride along s
3, //stride along t
&ctrlPoints[O1[O1[Oll
4, llorder along s domain
4, llorder along t domain
GL MAP2-VERTEX-31;
gl&d~urfacelp~urbl;

We can further enable lighting for the Nurbs surface, by defining lights and
material settings in the scene. We define our Nurbs surface to have a blueish
material and set up some default lights. In order to render the surface, we need
to enable automatic normal vector calculations for the surface by calling:

The complete code is given in Example7-3/Example7-3.cpp.We have also

put in a timer that moves the control points to simulate a wave animation on the
surface.
Subdivision Sutfaces
Nurbs have become the standard represcntation for complex smoothly varying
surfaces. However, a major drawback of Nurbs is the requirement that control
nets consist of a regular rectangular grid of control points. Subsequently, a Nurbs
surface can only represent limited surface topologies. Nurbs also they lack a local
rcfinement procedure of the model and surface discontinuities (places where we
want CO or C1 continuity only) cannot be locally introduced or controlled.
Subdivision surfaces-defined as the limit of an infinite refinement process-
overcome many of these deficiencies. For instance, the images above show an
initial control mesh and the mesh after onc refinement stcp, after two refincmcnts,
and at the limit of infinite refinement, respectively.

Fig.7.12: Subdividing a surface

Subdivision surfaces are easy to implement. They can model surfaces of
arbitrary topological type, and as shown above, the continuity of the surface can
be controlled locally. They arc becoming the dc-facto standard in the CG
industry to model complex surfaces. Refer to DER098 for more information on
subdivision surfaces.

CSG Models
CSG attempts to build objects by adding and subtracting basic shapes to generate
more complex shapes. For example if we takc a cube and subtract a sphere from
it, we could define a bathroom sink model as shown
Many more modeling techniques exist. Particle systems arc used hcavily to
model particle-based objects like fire or water. Another commonly used
tcchnique to get the model into thc computcr is to scan it in using a scanner. The
scanner digitizes the important features of the model and records these as data
points. Thcsc points can then be used to generate component polygons or be
approximated into a Nurbs surface. Each technique is usehl in its own way and
has its own drawbacks. You should pick thc tcchnique most suited to your
situation at hand.
 Fig.7.13: CSG to define a sink
And now, let's look a little bit more into some advanced techniques to render
photorealistic images.

7.2 Advanced Rendering Techniques

Shading techniques such as those discussed in Chapter 6 go a long way towards
creating colorful images. But they fall short of achieving photorealism. Real
objects cast shadows (shadowing), some of them reflect other objects
(reflection), and some of them allow light to pass partially or wholly through
them (transmission). There are ways to "cheat" such effects. We have seen a saw
a way to use texture mapping td render objects reflecting their environment.
There are even techniques to fake shadows, but none of them can take into
account all the various lighting effects found in the real world.
The interplay between various lighting effects is complex. One of the best
ways to take all the different lighting phenomena into account is by a technique
called ray tracing. Ray tracing can create stunningly photorealistic images (see
Color Plate 5). Its greatest
- drawback is that it is very slow and mathematically
very complex. It is usually combined with other rendering techniques to achieve
the desired effects.

Ray Tracing
Conceptually, ray tracing is a simple process. We simply need to trace the path
of all the light rays in the scene, bouncing them around along the path would
follow in real life (based on the physics of light we learned about earlier). We can
then combine the color of all the rays which strike our image plane to arrive at
the final color for each pixel. (Contrast this approach with that used in Chapter
6, where we only used rays emanating from a light source that got reflected and
directly bounced into the eye.) Fig. 7.14shows a scene of a room with a table, a
martini glass, a mirror, and two sets of lamps. Rays from the lights bounce
around the room to illuminate it.
 Fig.7.14: Light rays bouncing in a room
Let us trace the path that some of these rays would actually follow in real life.
Ray A from the lampshade, strikes the wall and is absorbed by the wall. Its
contribution to the scene ends here. Ray B hits the mirror and is reflected into
the image plane to finally enter the eye. We should see the image of the lamp at
this pixel. Ray C is transmitted through the glass, reflects from the table, and is
again reflected into the image plane. We should see the table color with the color
of the reflected image of the glass at this pixel. In general, the light rays leave
the light source and bounce around the scene. The various reflections and
refractions of the rays cause the color of the ray to change. Only the rays that hit
the image plane, however contribute to the image. The color of the ray when it
strikes the image plane will be the color contributed to the image. Some areas are
never lit because rays from the light are obstructed by another object, causing
shadows. Color Plate 5 shows a ray traced image of such a scene generated using
Maya. We have just been ray tracing; we have followed the path of the rays as
they bounce around the scene. More specifically, we have been forward ray
tracing, tracing the path of light starting from their origin. Forward ray tracing is
very inefficient, as most of the rays do not actually contribute to the image.
Tracing backward from the viewpoint to the light sources increases the
efficiency of the process by making sure that we only trace the paths of those
rays that finally reach the eye. In fact, ray tracing always refers to backward ray
tracing.
The process of backward ray tracing works as shown in Fig.7.15. For every
pixel of the image plane, a ray is drawn from the eye through the pixel and then
intersected with models in the scene. Each time a ray strikes a surface, additional
rays are drawn based on the physic of light. If the surface is reflective, a reflected
ray is generated. If the surface is transmissive, a transmitted ray is generated.
Some surfaces will generate both kinds of rays.
In Fig.7.15 we show one such ray striking the table. The table is a reflective
surface. We reflect this ray backward, to strike the martini glass. This ray will
produce the reflected image of the glass on the table surface. The glass is
transmittive as well as reflective, so two kinds of rays can be spawned at this
point, causing reflections in the glass as well as the "see through" appearance of
 Fig.7.15: Backward ray tracing
the glass. At every point of contact, a shadow ray is drawn from the point of
intersection to each light source in the scene. If the shadow ray strikes another
surface before hitting the light, the surface of intersection is in the shadow of the
obstructing model. In our case, the point of contact of the ray with the table is in
the shadow of the second glass from Light la. This will cause a shadow of the
second glass at this point.
Mathematically, all these rays with material characteristics of the models tell
the computer about the color and intensity for every pixel in the image. After a
few bounces, the rays are so weak they don't contribute much to the scene. To
avoid infinitely long computations, we usually we set a limit to how "deep" we
trace the rays. For example, we may set the trace to be only 3 levels deep. We
would then stop tracing the rays further. We shall actually generate ray-traced
images in the next chapter. Refer to [GLAS89] for more details on ray tracing
and how to implement your own ray tracer.

Radiosity
Simple recursive ray tracing has many unquestionable advantages. It does a great
job of modeling shadows, specular reflection, and (dispersionless) refractive
transparency.
Unfortunately, ray tracing also has a number disadvantages:
w Some ray tracing methods are extremely computationally expensive and
require much time to produce an image.
w The rendering time is greatly affected by the number of textures and
light sources and by the complexity of environment.
w Ray tracing is view dependent, which means that the all pixel values
have to be re-calculated for every new position of the observer.
To satisfjr demand for generating photorealistic images, without ray tracing's
imperfections, computer graphics researchers began investigating other
techniques for simulating light propagation. In the mid-1980s, radiosity, as this
technique is called, was finally developed.
The radiosity method of computer image generation has its basis in the field
of thermal heat transfer. Heat transfer theory describes radiation as the transfer
of energy from a surface when that surface has been thermally excited. This
 encompasses both surfaces which are basic emitters of energy, as with light
sources, and surfaces which receive energy from other surfaces and thus have
energy to transfer.
This "thermal radiation" theory can be used to describe the transfer of many
kinds of energy between surfaces, including light energy.
As in thermal heat transfer, the basic radiosity method for computer image
generation makes the assumption that surfaces are diffuse emitters and reflectors
of energy, emitting and reflecting energy uniformly over their entire area. It also
assumes that an equilibrium solution can be reached; that all of the energy in an
environment is accounted for, through absorption and reflection.
In an enclosed environment conservation of energy assures that the energy
leaving a given surface is incident upon other surfaces, thus a system of
equations (the radiosity equations) may be formed describing the
interdependence of the surface energies.
All that is needed to calculate these equations is the geometry of the scene,
the energy of the light sources (measured in radiosity units, energy per unit area
per unit time), and the diffuse reflectivity at every point on every surface in the
scene.
It should be noted the basic radiosity method is viewpoint independent: the
solution will be the same regardless of the viewpoint of the image and needs to
be calculated only once per scene.
Like ray tracing, radiosity equations produce stunning images. Refer to
[DUTR03] for the latest in radiosity technology. Radiosity is able to very
effectively handle a phenomenon called color bleeding: where one surface's
color seems to bleed onto neighboring surfaces. However, it operates at the cost
of high memory usage and also cannot account for specular highlights since it
assumes that every surface is a diffuse reflector. Probably, in the future, we shall
see techniques that combine ray tracing with radiosity to achieve even more
stunning effects.

Summary
In this chapter, we have learned some advanced techniques to model and render
CG scenes. We have used OpenGL to define and display cubic curves and
surfaces.
In the next chapter we introduce you to Maya. Using Maya, we will model a
3D scene and render it using ray tracing
Chapter 8
And Finally, Introducing
Maya

You may have realized how difficult it is to visualize 3D space in your mind or
on a piece of paper. You may be thinking of creating cool models, but you don't
know where to position the points to get the model of your dreams! Never fear
modeling software is here.
Modeling software was developed to aid in the 3D visualization process for
modeling. Most 3D software now have a toolkit for rendering and animation as
well.
The 3D software package we discuss in this book is called Maya. Maya has
been developed by a company called AliasIWavefront. Academy Award winning
Mayam3D is the world's most powerhl3D animation and visual effects software
for the creation of advanced digital content. Maya provides a comprehensive
suite of tools for your 3D content creation work, ranging from modeling to
animation and rendering to name but a few.
If you have access to Maya, great. Otherwise, you can download the personal
learning edition of Maya (Maya PLE, its free) from the Alias web site. Refer to
the Appendix on how to download and install Maya PLE.
In this chapter, you will lean the following:
The basics of working with Maya
rn Using Maya to model polygonal and Nurbs surfaces
Using Maya to define material properties
rn Rendering a scene using ray tracing
8.7 Maya Basics
Critical to learning any software program is an initial understanding of some
basic concepts: The UI of the software, the hot keys and terminology used by the
software, etc. Maya is a tremendously powerful toolkit. We cannot imagine
teaching you all that it can do in one chapter. However, with a few basics in
place, it will be easy for you to unleash its power.
For clarity, we use a number of conventions.
When you are instructed to select a menu item we use the following
convention:

w Menu > Command (For example, File > New Scene)

When you are instructed to select the option box for a particular menu item, we
use the following convention:
W Menu > Command > Option (For example, Create > NURBS >
Primitives > Sphere > 0 .

We use the same convention for selecting from the shelf or for any other
part of the Maya interface.

To start Maya
Do one of the following:
w Double-click the Maya icon on your desktop.
w (For Windows 2000 Professional) From the Windows Start menu, select
Programs > Alias>Maya (version number) > Maya (Complete, Unlimited or
Personal Learning Edition)
Let us first understand what we are seeing and how to use the interface. There
are a lot of items displayed in the Maya user interface. We won't look into all
these components right away. We will start with just the ones we need to create
our first few models. As we need to use more interface components, we will
introduce them to you.
The User lnterface
The user interface refers to everything that the Maya user sees and operates
within Maya. The menus, icons, scene views, windows, and panels comprise the
user interface.
Menu BY Slllw Unr She1 WWow Channel Box

Fig. 8.1: The Maya User Interface

When you start Maya for the first time, the workspace displays a perspective
projection into the 3D world. This window is called the perspective window. The
window displays a grid (also called the ground) in the xz plane. The intersection
of the two heavy lines( the x and z axis) is our origin. The y-axis points upwards.
You should be familiar with these concepts by now.

Fig. 8.2: The coordinate system in Maya

Maya labels the x-, y-, and z-axes with a color scheme: red for x, green for y,
and blue for z. The color scheme is a visual aid used extensively in the modeling
of objects.
The Main Menu
Maya has support for modeling, animation, rendering, and various other CG
related activities. Each activity corresponds to a module within Maya. Modules
are a method for grouping related features and tools of the task you are
performing.
The Main Menu bar at the top displays the menu set for the module within
which you are currently working. You can switch between menu sets by
choosing the appropriate module from the menu selector. Let us create a
primitive 3D object from the Modeling menu set.
Choose the Modeling module.
Menu Seleclor
I

From the main menu bar

Select Create > Polygon Primitives > Sphere
Maya creates a 3D sphere object and places it at the center (origin) of the
Maya workspace. This primitive is defined by a polygon mesh. The model
displays in a green color.
We learned earlier that a Nurbs surface (by definition) needs fewer points t
han a polygonal surface to represent a smooth surface. To create the same
primitive using a Nurbs surface:
Select Cueate>Nurbs Primitives>Sphere
A sphere primitive is created which is defined by a Nurbs surface. The new
model shows up highlighted in green on the screen. The previous model now
turns blue.

The Status Line

The Status line is located directly below the Main Menu bar. It contains a variety
of selections, most of which are used while modeling or working with objects
within Maya. Many of the Status line items are represented by a graphical icon.
The icons save space in the Maya interface and allow for quick access to tools
used most often.

You've already learned the first item on the Status line: the Menu Selector
used to select between menu sets.
The second group of circled icons relates to the scene and is used to create,
open, and save your Maya scenes.
We will consider the other items shortly

Shelf
The Shelf is located directly below the Status line. The Maya Shelf is useful for
storing tools and items that you use frequently use. Some of the Shelf items pre-
configured for your use, tabbed into sections for convenience.

To create an object using the Shelf

From the Shelf, select the Surfaces tab in order to view the tools located
on that shelf. This tab enables you to create Nurbs based surfaces.

Select the cylinder icon located on the left end of the shelf by clicking on
it
Maya creates a cylinder primitive object as a Nurbs surface and places it at
the center of the scene. By our convention, from now on, the above command
will be specified as:
Select Surfaces>Cylinder from the Shelf.
(The same command can also be driven by the menu bar, as we saw earlier.)
In your scene, view the wire-frame outline of the spheres you created earlier.
They have changed color to navy blue, and the cylinder is displayed in a bright
green color. The cylinder is now the selected object (the sphere is no longer
selected). In Maya, when the object is displayed like this, we refer to it as being
selected or active.
Selection of an object or a component indicates to Maya that this particular
item is to be affected by the tool or action you will subsequently choose. As you
work with Maya, you will be selecting and deselecting items a lot.
To save your scene, choose the Save icon on the status line. You will be
prompted to enter a name for your 3D scene. Enter a name and save your world
into the Maya proprietary model file. You can always load it back up whenever
you want.
You should have the basic idea now to begin!
In this chapter, we will use Maya to model and compose our 3D world and
finally render the scene as a ray-traced image, as shown in ColorPlate 5.
188 8 2 MODEL
ING 3D OBJECTS

8.2 Modeling 3 0 objects

We shall use Maya to model the objects shown in Fig.8.3

Table Mirror

Lamp I Lamp 2 Martini Glass

Fig.8.3: The objects to model

When you first begin to construct the models of our world, it is useful to
know the size of the objects relative to each other. You can do this by drawing a
rough composition of your scene and taking approximate measurements of the
objects within it. This helps to avoid scaling issues later on when we compose
the scene. (Of course, you can always change the sizes of the models at any
time.) The basic dimensions of our objects are also shown in Fig.8.3.

The Table:
Let us start by creating the table. The table, as shown in Fig.8.3, can be thought
of as a rectangular cube (the base) with thinner but longer cubes as the feet. The
dimensions for this model are as follows:
15 by 2 by 7 units for the base.
1 by 6 by 1 unit for the legs.
Start Maya, if it is not running already.
Start a new scene by clicking the Create New Scene icon on the status line.
We will define the components of the table to be polygonal primitives, since we
don't need the complexity of a Nurbs surface. First, we define the base of the
table.

The Base
1. Choose the PolygonXube option from the shelf.
A unit cube will display in the perspective window. We want to scale this
cube to the dimensions specified. Scaling is easier to define in orthographic
windows, where we can see the exact x,y,z locations of the models.

Layout Shortcuts

I
The tool-box is a UI component located on the left hand side of the Maya

View Shortcut

user interface. The Quick Layout buttons on the lower half of the tool-box
provide a shortcut for you to select a different window panel or switch to another
layout.
2. From the tool-box, click the Four View layout shortcut.
The workspace changes to a four-view layout. The three new windows define
the three orthographic projection views along the x-, y- and z-axis (See Chapter 5).
&- -+I - -- , i ; ~ ~ % % + x , ? ~ ? ~

.*, A ' . q , ,a. - '"" ) i . ' ' C

-- --

-. I :
Side V~ew '

Fig.8.4: The four views: top, front, side and perpective

190 8.2 MODELING
3D OBJECTS

They are referred to as the top view, side view and front view windows. These
windows are an invaluable tool in visualizing and designing the object from
different angles. The perspective view is located in the top right corner, and the
other views show the object from the top, front and side.
Each view has a grid defining the grid line markings along the axis. These
lines are like the lines of a ruler and are an invaluable tool in determining the size
of models. Let us set the markings so we can size our models within the same
units.
3. Choose Display>Grid> O from the menu bar
4. Set the Grid Lines Every option to be every 1 unit.
5. Set the Subdivisions options also to 1.
These steps will set up the grid lines to be drawn every 1 unit, with no
subdivisions in between. Every grid line corresponds to one unit in our world.
We will use these lines to size our models appropriately.
To size the table's width and length, we need to size it along the x-,y- and z-
axes. To do this, we will need to use the Scale transformation tool located in the
Tool Box.

The upper half of the Tool Box contains the tools for transforming objects
(selection, move, rotate, scale) within Maya. (When you move your mouse
cursor over any icon, the name of the tool appears next to the mouse cursor,
indicating what action you are going to commit).
Before you can transform an object, you must ensure it is selected. You can
select an object by clicking on it directly, or by dragging a rectangular bounding
box around some portion of the object to indicate what you want selected. To
deselect an object, you simply click somewhere off of the selected object.

6. If it isn't already selected, then click on the cube with the left mouse
button in order to select it.
7. Click on the Scale tool from the toolbox.

A scale manipulator icon appears over the primitive cylinder in the scene
view.
The Scale Tool Manipulator has handles that point in the direction of the
three axes. They are colored red, green, and blue, and control the direction of the
scale along the axes.
 When you click a specific handle, it indicates that the move is constrained to
that particular axis direction.

8. In the top view window, drag the red x manipulator handle to scale the
object along the x-axis. As you drag the manipulator, you will see the outline
of the shape scaling up. Scale it enough so that it occupies 15 boxes along

Fig.8.5: Scaling the table

the x-axis. This sets the length of the base to be 15 units.

In the top view window, drag the yellow z manipulator handle to scale the
object along the z-axis. Scale it so that the object occupies 7 boxes along the
z-direction, making it 7 units long.
To change the height of the base, click on the green y manipulator handle in
the side view window, and drag it so the object occupies 2 boxes along the
y-axis.
There! We are done creating the base of the table.
But wait! We still need to create the legs.

The Legs
The legs of the table can be defined similar to the base.
1. Choose the Polygon>Cube option from the shelf.
2. Scale this new cube by Sx=l, Sy = 6 Sz = 1 or (1,6,1). (Remember to select
the new cube before applying the scaling transform.)
Now we need to move this leg to its correct position.
3. Select the Move Tool from the Toolbox. A move manipulator icon appears
over the cube.
The Move Tool Manipulator has handles that control the direction of the
movement along an axis.
When you click a specific handle, it indicates that the move is constrained to that
particular axis direction.
4. In the front view window, drag the green y manipulator handle to move
the leg downwards in the y direction. Move it downwards enough so that
192 8 2 MODELING
3D OBJECTS

Fig.8.6: Attaching legs to the table

the top of the leg touches the bottom of the base as shown in Fig.8.6
5. Drag the red x manipulator and finally the blue z manipulator in the top
view window in order to move the leg to the right hand (front) corner of
the table.

We need three more legs. We can create them by duplicating the leg we already
created. The shortcut to duplicating an object is the hot key: Ctrl-D (i.e., pressing
the Ctrl and the D buttons of the keyboard together)
6. Select the leg. Hit Ctrl-D. This will create a copy of the selected object.
7. Choose the Move transform and move the newly created leg to the left
comer. Repeat the process for the back two legs.

All Together now

We have a table! We should group the components of the table under a hierarchy,
so we can transform the table as a single model. To do this, we perform the
following steps
1. Simultaneously select all the objects that comprise the table by doing one
of the following:
0 With your left mouse button, shift-click each object in turn until all the
objects are selected in the scene view (that is, click on each object
while holding the shift key down).
With your left mouse button, drag one large bounding box around the
five objects using an orthographic view.
2. From the main menu bar, select Edit > Group > O .
This step will popup a new window called the grouping window. The
grouping window allows you to group the selected objects under a new parent
node (the Null Node we saw in Chapter 5). The pivot point for the parent node
can be placed at either the world origin (0, 0, 0) or at the center of the grouped
object. We will set it to be at the center of the grouped object. In the grouping
window:

3.Set the pivot point option to be Center.

4. Set the Group Under option to be Parent.
5. Click the Group button.
The components will be grouped together as leaf nodes under one parent.
Maya offers different selection modes: Hierarchy, Object ,and Component.
You use these modes in order to select only the types of items you want.
The icons for the three modes appear on the Status Line.

When you first start Maya, the default selection mode is set to Objects. This
means we select individually defined components. Now that we have grouped
our components under a hierarchy, we want to select the entire hierarchy. To do
this
6. Change the selection mode to Hierarchy by clicking on the Hierarchy
selection mode on the Status line.
7. Now if you click on any component of the table, the entire hierarchy o f t
he table model is picked.
8. Save your work by choosing the Save Scene icon from the status line.
9. Save your model in a file called Table. Maya saves the model in its
proprietary format.
Whew. One model down, four to go. By now you must have started to
appreciate the power of modeling software such as Maya. But the h n is yet to
begin!

The Mirror
The steps to create the mirror are similar to those we followed when modeling
the table. The mirror can be modeled using two cubes with the dimensions
shown in Fig.8.3. The reflective (mirror) should be slightly in front of the frame.
1. If Maya is not running, start it again.
2. Start a new scene by selecting the Create New Scene icon on the status
line.
3. Create a polygonal cube.
We wish to scale the cube so that it has the dimensions (15,10,1). You can use
the scale manipulator we saw when creating the table rr we can use the Channels
editor. The Channels editor shows up on the right-hand side of the Maya
interface and displays transformation information for the currently selected
object.
If you don't see the Channel editor click on the ShowIHide editor for the
Channel box on the status line.
bns
Snap ~ o d a ShOwMido Mar %om

4. Make sure your cube is selected.

5. In the Channels editor, set the value of ScaleX (Sx in our terminology)
to be 15, ScaleY=lO, and ScaleZ= 1.
6. Create another cube.
7. Set this cube to have ScaleX = 13 and ScaleY = 9.
8. Translate the newly created cube to be slightly ahead of the other cube
by setting the value of TranslateZ (Tz) to be 0.3.
9. Group the objects by selecting them both and hitting the hot key to
group objects--Ctrl-G -a shortcut. (If you had restarted Maya from the
last model exercise, you may have to reset the grouping options by
using Edit >Group>O)
At this point, let us see how we can find out about the items in the scene.
10. Choose Window > Outliner from the menu bar.
The Outliner window pops up. The window displays a list of the items in the

1=.---
& p . ~Show
----
Help

scene: including the windows and lights being used. It is very useful for
grouping, setting names for models, choosing models in the scene, etc.
Your two cubes will show up as pCubel and pCube2 (p is for polygonal), or
something similar.
 11. You can rename the newly created parent as Mirror by double clicking
on the name of the group and entering the desired name.
12. Save your scene in a file called Mirror.

The Martini Glass

Many objects in our world have a perfect symmetry about an axis. Objects such
as a vase or a glass are typical models that display such a symmetry. These kind
of objects can be easily constructed by revolving a master contour about the axis

Fig.8.7: A model built by revolving a master contour about an axis.

of symmetry, as shown in Fig.8.7.
Our martini glass is a perfect candidate for a surface of revolution. Its axis of
symmetry is along the y-axis. Let us see how to create this model using Maya.
1. Start a new scene in Maya.
2. Click on the Four View shortcut.
3. To enlarge the front view, click the mouse cursor inside the front-view
window and tap the spacebar of your keyboard.
The workspace changes to a single-view layout with the front view in an
enlarged view.
4. Choose Display>Grid Ofrom the menu bar.
5. Set the GridLines Every to be 1 unit and the subdivisions to be 5. This
means we have five sub divisional lines drawn for every 1 unit.
6. Since the glass has a maximum extents of only 3 units, zoom into the
window by pressing the Alt key and dragging the mouse with the right mouse
button held down. Zoom in so you can see around 20 grid lines.
To draw the curve defining the martini glass contour, we will define a Nurbs
curve.
7. To define a Nurbs curve, choose Create > CVCuwe Tool from the menu bar.
8. In the front-view window, click your mouse to define vertices at the
numbered positions as shown in the Fig.8.8.
This will define the control vertices for our curve. Make sure the first and last
positions (Control point 1 and 14) are located on the grid's y-axis (the thickest
vertical line of the front view's grid). The first CV (Point 1) is shown as a square
and the u icon depicts the tangent of the curve from the starting point.
9. Click two times in the same spot for positions 12 and 13. This action leads
to the double-knot situation we saw in Chapter 7 and is useful when creating sharper
corners. Three CVs at the same point would create a linear comer (three knots).
10. Press Enter to complete the curve creation.
 Fig.8.8: The master contour of the martini glass

If you find you need to edit the curve follow the steps below
I. Select the curve.
2. Click down on thc right mousc button until a ncw window called the
Marking Menu pops up.
The marking menu is used for context- related actions. The way to use it is to
click the right mouse down till the menu shows up, and then drag the mouse to
select the item desired.
3. Do not release the mouse. Drag the mouse to the menu item marked
Control vertices to highlight this item.
4. Release the mouse button. You will now be able to see the CVs of
your curve.
5. If you select the Move Tool and then click on a CV, you will see the
move manipulator. Use the manipulator to move the CV around as
desired.
6. Repeat step 5 until all thc CVs are positioned appropriately.

I I . Using the marking menu, select the Object Mode item. The CVs
disappear.
12. Switch back to the four-window layout by clicking on the Four View
layout icon in the toolbox.
13. Select the curve by clicking on it.
14. To revolve this curve, select Surfaces>Revolve from the menu bar. You
will see the revolved surface appear in the perspective window.
15. Click in the perspective window.
16. You can zoom in by pressing the Alt key and dragging the mousc with
the right-mouse button held down. Dragging the left-hand mouse button
with thc Alt key held down will enable you to view thc glass from
different angles. Pressing the Alt key and dragging with the middle
mouse button held down will cnablc you to move the sccnc up and down.
17. Click in the perspective view, press the keyboard key 5 (a shortcut for
Shading>Smooth Shade All option in the menu bar in the perspective window).
This displays the martini glass as a shaded surface. Spin the view around so you
can view your model from all angles. Select the keyboard key 4 to display the
model in wire-frame mode again.
18. If you are not satisfied with your model, choose Display>Outliner. The
outliner window will enable you to select just the Nurbs curve you had
first created.
19. Follow steps I to 6 to modify this curve.
20. Your surface of revolution will change its shape as you move the CVs
around.
21. Save your scene into a file called MartiniGlass.

Stirrer
Do not create a new scene. We will create a stirrer with the martini glass as our
guide.
1. Select the martini glass (and the curve used to create it).
2. Select Display>Object Display>Template from the menu bar. This will
template the martini glass. A templated model still displays the wire-frame hull
of the mode, but you will not be able to easily select it.
We wish to create the stirrer as a cylinder with radius=0.05 and height = 2. You
can use the Polygon>Cylinder option from the Shelf as w had done earlier.

3. An easier way is to choose Create>Polygon Primitives>Cylinder> from

the menu bar.
4. A popup a window will allow you to set the radius and length of the
cylinder.
5. Set radius = 0.05 and height =2. Then click on the Create button. This will
create the cylinder with the specified dimensions.
Now, we will put a sphere on the the tip of the stirrer.
6. Choose Create>Polygon Primitives>Sphere>O and set the radius to be
0.075.
7. Click Create to create the sphere.
8. Move the sphere to the tip of the stirrer by setting its TranslateY = 1.9.
Let us also create an olive on the stirrer.
9. Create a sphere with radius = 0.3. We wish to scale it along x and z to give
it an elongated shape.
 10. In the Channels editor (with the olive picked) set its ScaleX and ScaleZ
to be 0.7.
I I . Group the three objects (the stirrer, the olive, and the tip) under one
parent.
12. Set the selection mode to be hierarchy and select the stirrer.
Let us rotate the stirrer so that it resides appropriately in the glass.
13. With the stirrer selected, choose the Rotate tool from the tool-box.
A rotate manipulator icon appears in the scene view.

Rotate
manipulator

The Rotate Tool Manipulator consists of three rings (handles), plus a virtual
sphere enclosed by the rings. The handles are colored red, green, and blue based
on their fimction related to the x-, y- and z- axes and control the direction of the
rotation around an axis.
14. In the front-view window, drag the blue Z manipulator ring to rotate the
stirrer so it is aligned with the edge of the glass.
You are rotating the cylinder around its z-axis. What is the pivot point?
Alternately, you can also set the RotateZ (Rz) value for the group to be about 6.3
in the Channels editor.
15. You may have to translate the stirrer further to place it within the glass.
16. To define the drink inside the glass, create a polygonal cone primitive
with a radius of 1 unit and a height of 1 unit also.
17. Rotate the cone by setting its RotateZ = 180
18. Position it within the glass by settings its TranslateY = 0.7.
19. Click Edit>Select All from the menu bar.
20. Click Display>ObjectDisplay>Untemplate from the menu bar to un-
template the glass.
21. View your creation in the perspective window by hitting 5 to see the
objects as solid.
22. Choose Window > Outliner.
The Outliner appears and displays a list of the items in the scene.
23. In the Outliner, select the curve you revolved and delete it, so you will
not accidentally modify it later on.
24. Group all the objects under one parent.
25. Rename the group to martiniglass.
26. Save your work back to the file called MartiniGlass.

Lamp 1
Lamp1 has a cylindrical stand with dimensions as shown in Fig.8.3.

1. Start a new scene in Maya.

2. Reset the grid so the gridlines are drawn only once every unit.
3. Create the stand by choosing Create>Polygon Prirnitive>Cylinder>O
4. Set the radius to be 0.3 and the height to be 20. Click Create.
The shade can be modeled as a cylinder, that has been scaled on the top to define
a conical look. We will define it to be a Nurbs surface to get a smoother look.
 5. Choose Create>Nurbs Primitive>Cylinder> O and set the radius = 3.
and height = 4.
6. Select the cylinder and select Control vertex (CV)from the marking
menu. You will see and be able to pick the CVs associated with this Nurbs
surface
7. In the front-view window, select the top ring of CVs by dragging a box
around them.
8. Select the Scale tool from the tool box.
9. In the top-view window, scale the ring of vertices down along x and z.
10. Do the same for the second ring of CVs, but scale it less than the first
ring.
We are munging the vertices defining our Nurbs surface. Recall that this
affects surface patches local to the vertices. The front view of the lamp shade
should now appear as shown in Figure above.
11. View your model in the perspective window.
12. Group and name your object lamp 1.
13. Save you work in a file called Lampl.

1. Start a new scene

2. Create the stand for lamp2 as a polygonal cylinder with height = 20 and
radius =0.3
3. The lampshades will be Nurbs cones. Select the Create>Nurbs
Primitive> Cone>O.
4. Set radius = 1.5 and height = 3. Make sure the Caps option is set to
None.
We will create a cap for the cone, but we wish it to have a different material
than the shade. This will serve as the light beam shining from inside the shade.
For this, we will cap the cone with a polygonal cylinder.
5. Select Create>Polygon Primitives>Cylinder>O and set radius =1.3
and height = 0.1.
6. Translate the cylinder by settings its TranslateY= 0.25, so that it caps

Fig.8.10: Modeling the second lampshade

 the cone.
7. Group the cone and the cylinder, and call the group Shade.
8. Make sure the selection mode is set to hierarchy.
9. Select the shade and rotate it by setting RotateZ = 40.
10. Move and translate the shade by settling TranlsateX = 2 and TranslateY
= 3.1
11. Lamp2 has two shades. Duplicate the shade created by hitting Ctrl-D.
12. Set the new shade to have its transforms defined as RotateZ = -120,
Translatex = -2.5, TranslateY = 1.8, so that it is on the other side of the
lamp.
13. Group the three objects under one parent called lamp2.
14. Save the scene as Lamp2.

8.3 Applying Surface Materials

Let us apply surface materials to our objects, so that they start to look more
appealing. To ensure the steps work as described, select Rendering from the
menu selector. Tab to the Rendering tab on the Shelf. This will enable the
shortcuts to the common rendering tools.

We shall start by defining the materials of the two lamps first, since these are
the simplest. We will then define the material of the martini glass which needs
transparency defined for the glass surface; the mirror, which needs reflectivity
settings; and finally the table, which needs a texture setting for its wood.
By default, Maya assigns a default material called lambert to all models
created. This default material is grey in color and has diffuse-only reflection
defined. To assign new materials to objects, we need to create new material
attributes.

Lamp 1
1. In Maya, click on the Open a scene icon from the Status line.
2. Type in the name of the model file where you saved Lampl. This will
open up your Lampl model.
3. Set the selection mode to Object and select only the lamp stand.
The Lamp Stand
Let us first define the material for lamp stand. We want the stand to be a
reflective metallic material so we will need to create a new material for this
object.
1. Right-click the stand to select it.
2. Select Rendering>Assign a New Phong Material icon from the shelf.
This option will create a new material and assign it to the lamp stand. Phong
shaders includes all three components of the illumination model: ambient,
diffuse, and specular components. A Phong material will give a surface a shiny
look.
The Attributes editor will appear on the right hand side of the Maya UI. The
editor will display a sphere rendered with the current material attributes. If you
don't see the Attributes editor, click the ShowIHide attributes editors from the
status line (it is to the left of the Channels editor button).
Below the sphere on the Attributes editor, is the menu to modify the attribute
settings of this material.
3. Click in the text box that defines the name of the material and change it
to Chrome so that we can identify this material for use later on in the
project. The name of the material is at the topmost text box of the editor.

-.- -
L
i-

-- , . .
c-i-
-7,
-ma..---. t

-I.-_
--• ?;

t
I- I Y- l -1. i

4. In the Common Material Attributes section click inside the color box to
the right of Color. This will open the Color Chooser window.
5. The Color Chooser window displays a color wheel. You can click inside
the color wheel (hexagon) and drag the mouse pointer to change the
color of your material.

6. You can also set the color by specifying the RGB components. Change
 the slider display from HSV to RGB as shown, and set the RGB values
as R=0.96, G=0.96, B=0.5-a yellowish color for our chrome stand.
7. Click Accept to choose the new color and close the Color Chooser.
8. You can change the intensity of the diffuse (and ambient) reflectivity of
your material by dragging the Diffuse Color (or Ambient Color) slider
to its right. Do so to increase the color brightness.
9. You can change the specular settings of your material by using the
Specular Shading section in the attributes editor. The Specular Color
option changes the specular reflectivity of the material (and its
intensity), and the Cosine Power slider changes the size of the shiny
highlights.

10. Ignore all the other attributes for now. Click in the Perspective window.
11. Select the IPR Render Current Frame icon from the Shelf.

This will render an image of the currently selected window. Since you
clicked in the perspective window earlier, this will render an image of your lamp
as seen in the perspective window.
12. You can change the perspective camera (by hitting Alt and Mouse
right/middle/left) to view the lamp from different directions. Hitting the
IPR button will render the lamp from the new viewpoint.
13. Change your material until you are satisfied with your render.
14. If you lose the material editor, choose your lamp stand again and choose
the Show shading group attribute editor icon from the Shelf. You can
then tab to the Chrome material in the Attributes editor

The lampshade
The lampshade is made of cloth and has a diffuse material. We will create a new
Lambert material for the shade. Lambert shaders have only ambient and diffuse
components and hence do not look very shiny-Perfect for a cloth.
1. Select the Rendering>Assign a new Lambert material icon from the Shelf.
 2. Set the name of this material to be shade, and set its color to be R=0.9,
G=0.5 B=O-an orangish tinge
When you see a lamp shade in real life, it seems to glow from the light within it.
We can achieve this effect by giving the shade a bit of emissive color.
3. To do this, drag the Incandescence slider, under the Common Materials
Attributes section, to the right. The Incandescence of a material defines
the emissive nature of the material.
4. Render till you are pleased with the results!
5 . All done with lamp1 ! Save your scene back to avoid losing your work.

1. Select the Open a new scene option and open the file Lamp2.
2. Set the selection mode to be Object mode.
3. Select the stand and assign a new Phong material to it.
4. Set the name of the material to be metal.
We shall assign a metallic-looking material to the lamp and the lamp shades.
5. Assign this material to have a whitish color. RGB = (0.8, 0.8, 0.8).
6. Under the Specular shading options, set the specular color to be the
same color.
If you have looked carefully at a metallic object such as a spoon, you may
have noticed that you can see reflections in it. Metallic objects reflect their
environment. Maya enables materials to reflect their environment by the value of
the Reflectivity attribute. The slider to control this attribute ranges from 0 (no
reflections) to 1 (clear reflections). The default value is 0.5. Reflectivity values
for common surface materials are car paint (0.4), glass (0.7), mirror (I), metal
(0.8).
7. Set the reflectivity value (under Specular Shading options) to have a
value of 0.8.
8. You will not see the reflections in the IPR render. When we ray-trace
this object, reflections in the object will give it its true metallic look.
9. Pick the cones of the shade and set them to have the metal attributes.
You can do this by selecting MateriaWAssign Existing MateriaWMetal from
the marking menu.
10. Pick the cylindrical caps of the shade.
We want the lamp shade to look like it's glowing-as if the light is coming out of
this object.
11. Select Rendering>Assign a new lambert material icon from the shelf.
12. Set the name of this material to be shade2, and set its color to be R=0.9,
G=0.9 B=O-a yellowish tinge.
13. Drag the Incandescence slider to the right. This component will change
the emissive color of the material.
14. Render.
15. Redefine the materials until you are pleased with the results!
16. Save your scene.
Martini Glass
1. Load up the martini glass.
2. Pick the revolved glass object.
3. We need to assign this object a glass material. Glass is transparent, but
it still displays specular highlights and some reflectivity.
4. Assign the glass a new Phong shader and call this shader, glass.
5. Leave the color of the glass as is.
6. Glass is transparent: we can see through it. To set the transparency of
the material, slide the Transparency slider. You can slide it to the right
to get a nice transparent glassy look- but don't drag it all the way to the
right, or your object will become invisible!
7. Under the Specular Shading options, we set the Reflectivity of the glass
to be 0.2. This will make the glass show very faint reflections.

The cone inside the glass is made of water. Water is also transparent, but it is
usually perceived as blue in color.
8. Pick the cone inside the glass. We will make this object be a water
material.
9. Assign it a Phong shader.
10. Set the color of the water to a bright blue. RGB= (0,0.9,1)
11. Slide the Transparency slider to get a nice transparent water look.
12. Again, we made the water nonreflective, although you can define a
small amount of reflectivity to it if you like.
13. Set the stirrer to also have the shader glass, by selecting
Materials>Assign Existing Material>glass from the marking menu.
14. If you lose your material, remember. You can always pick the model
and click on the Show shading group attribute editor icon from the
Shelf.
15. Create a new green-colored Lambert shader, with RGB=(0,0.5,0) for
the olive.
16. Save your scene to avoid losing your changes.

Mirror
1. Load the mirror.
 2. Select the outside cube-the frame of the mirror.
3. Assign it a new lambert shader.
4. Set the color of the shader to be magenta: RGB = (0.8, 0,0.2)
5. The mirror itself has a black color, and is completely reflective.,
6. Select the mirror.
7. Assign a new Blinn shader to it, and call the material mirror.
8. Set the color of the mirror shader to be black and the diffuse color also
to be black. The mirror itself dose not have any color. It merely reflects
its surroundings.
9. In the Specular Shading section, set the specular color of the mirror to
be 1 by sliding the slider all the way to the right.
10. Set the reflectivity of the mirror to be 1. The mirror reflects everything!
11. Save your scene

Table
1. Load the table model.
2. Pick the base of the table.
3. Assign it a new Phong material. Name the shader wood.
4. Click the checker icon next to the Color slider. This is a quick way of
applying a texture.

A window called the Create Render Node window pops up and displays a list of
the textures that you can apply. The texture names and their icons help you to
decide which texture is appropriate for the effect you want.
iGx-G- :----.
- -

Maya has both 2D textures and 3D textures. We saw 2D textures in Chapter 7.

They are essentially 2D images that you can map onto your object using a planar
mapping or by shrink-wrapping it onto the surface. In Maya, a 3D texture is
essentially a cube, with texture information on all sides of the cube. The texture
is then mapped onto the object using the cube mapping we saw in Chapter 7,
making the object appear as if it were carved from a block of the material
selected. We shall assign a 3D wood texture to our table.
5. Under the 3D textures option, choose the wood texture.

6. A green box indicating the position of the texture cube will be displayed
in the windows.
7 . The various parameters to define the values for this texture will appear
in the attributes menu.

8. Under Wood Attributes, click on the filler color. In the color selector, set
the RGB color of the Filler to be R=0.5, G= 0.26, B = 0.0 to get a rich
brown color, which is what oak wood normally looks like.
9. Under Effects, turn on the Wrap radio button, so the texture will repeat
itself. .
10. Render the table.
You will notice that the wood grain is closely packed. We need to scale the texture
box up to scale out the wood grain. Also, the texture needs to be rotated so the texture
cube is mapped correctly. We will need to rotate the texture box to achieve this.
11. Click on the place3DTexturel tab in the Attributes menu.
 This tab will enable you to position and scale the texture appropriately.
Essentially, you will be applying transforms in the texture space.
12. Set RotateX =-90 and RotateY = 90
13. Set ScaleX =2, ScaleY= 4 and ScaleZ = 1.

You will see the green texture box now engulfing more of the table.
14. Render.
15. You should see the tabletop rendered with a nicely defined wood
texture.
16. Assign the wood texture to each of the legs by selecting each one and
choosing MateriaWAssign Existing Materials> Wood from the
Marking menu.
In the Attributes menu, you can navigate to the wood shader by
selecting any object in the scene which has the wood shader defined.
You can navigate fiom the wood shader to the wood texture by clicking
the icon next to the Color slider.
When you render the image, the legs will not have the texture mapped onto it.
This is because the texture bounding box does not engulf the legs.
17. Scale up the wood texture placement by setting ScaleZ = 7.
18. Render.
19. Save your scene.

8.4 Composing the World

Let us now put all our models into the world to compose the scene as in Color
Plate.

If you have closed Maya, restart it. Open a new scene.

Load up the Table model.
Choose File>Import from the Menu.
Choose the file: Lamp2. This imports the lamp2 model into our scene.
Set the selection mode to be hierarchy.
Pick the lamp2 object.
Click on the Channels attribute box, and set the transforms for lamp2
to be
TranslateX = -2.7, TranslateY = 5.8, TranslateZ = -2.25
RotateX=O, RotateY = - 15, RotateZ = 0
Import the martini glass.
The glass should rest on the base of the table. Set its translations to be
TranslateX = 3.59, TranslateY = 0, TranslateZ = .472
 10. Duplicate the martini glass by hitting Ctrl-D
11. We want this second glass to sit beside the first glass. Set the
transforms for the second martini glass to be
TranslateX = 0.46, TranslateY = 2.5, TranslateZ = 0.5; RotateY =
158
12. Import the mirror. It resides behind the table, hanging on a (as yet
imaginary) wall. Set its transforms to be
TranslateX = -0.7, TranslateY = 5.437, TranslateZ= -5.125
In the perspective window, you will be able to see your world slowly being
defined. You can change the perspective camera to a camera setting that pleases
you. (Do this by hitting Alt and dragging the mouse with one of the buttons
pressed down)
13. Import Lampl. Lamp 1 will be positioned behind our camera, and it
will only be seen as a reflection in the mirror.
14. Zoom out in the top view and position the lamp in the front left hand
corner.
TranslateX = -10, TranslateZ = 23
We wish to define our 3D world within the confines of a 3D room. We need
walls for the room that we are position in!
15. Create a new unit cube.
16. Scale it to ScaleX = 35, ScaleY=35, ScaleZ = 35.
17. Position it as shown in Fig.8.11 by setting its translations as
TranslateX = 2, TranslateY = 11, TranslateZ = 11.
Assign a new lambert shader to the wall with a yellowish color.

Fig.8.11: World being defined in Maya

Click in the perspective window to select it and render your scene.

Voia! A 3D world. The scene looks good - but the lights are still default
lights. (If you are using the Learning Edition of Maya, then the render will have
the Alias logo over it as a watermark).
 We still need to define lights to define the mood of our world, and to render using
ray tracing so we can see reflections and shadows. Save your scene in a file called
world. We shall use this scene again in Chapter 11, when we animate the scene.

8.5 Lighting the scene

When you open a scene, Maya lights it with default lighting provided by an
imaginary light fixture.
Once you define your own lights, Maya will render using your defined lights
instead. Maya supports ambient light, directional lights (like sunlight), spotlights
(useful for cone lights), and point lights, to name a few.
Let us define lighting for the current scene.

Ambient Lighting
First we define some ambient lighting so that nothing in our scene is rendered
black.
1. Choose the Rendering>Ambient Light icon from the Shelf.

2. The attributes menu for the newly created light will show up.
3. Under the Ambient Light Attributes section, set the intensity ofthe light to be 0.2.

Light from Lamp1

We want to simulate lights from the lamps that we have modeled. Let us first
consider lamp 1.A point light is a good light to simulate light bulbs, which emit light
in all directions. We shall create a point light and position it at the center of lampl.
 1. Choose the Rendering>Point Light icon from the shelf.
2. A point light icon will appear at the center of the grid. You can move the
light just as you move objects. Translate the point light to be located where the
shade of lamp1 is.
3. In the Attributes editor, set the intensity of this light to be 0.3.
4. Turn off the Emit Specular radio button so that this light has only a difise
component.
5. Render your scene.
Your scene should look very moody now. We need to define the mood to be
a little more upbeat by adding some more lights!

Light from Lamp2

Lamp2 can also be thought of as simulating a light bulb. However, since it has
metallic shades, the light is not spread in all directions; it is restricted to a cone
defined by the shade. Spotlights create a cone of light and will be ideal to define
the lights of lamp2.

1. Select the spotlight icon from the rendering shelf..

2. This creates a spotlight. An icon for this light will display at the center
of the grid. The light points toward the back wall.
3 If you look at the spotlight icon from several angles, you'll notice that
it is shaped like a cone with an arrow pointing out of it.
4. The cone symbolizes the beam light that gradually widens with
distance. The arrow symbolizes the direction of the beam.
5. With the spotlight selected, select Modifi > Transformation Tools >
Show Manipulator Tool.from the Main menu. This tool provides two
manipulators that you can move to position and aim the light precisely.

Fig.8.12: Setting up spotlights

 6. The look-at point specifies where the light focuses. The eye point
defines the position of the light source. All types of lights have an eye
point, but not necessarily a look-at point.
7. Move the look-at point to the top of the table, near the wine glasses.
8. Move the eye point roughly to the center of the shade cylinder, as
shown in Fig.8.12. (We cheated a little so that our lighting effect could
be what we wanted! ! Actually the light should be placed inside the
shade itself.)
9. In the Attributes menu, set the color of the light to be slightly yellow in
order to simulate real bulbs in housing lights. RGB = (1,1,0.7).
10. The spotlight's default Cone Angle is 40 degrees. In the Attribute editor,
set the value of the Cone Angle to 20.
11. Render your scene. The circle of light is very narrow.
12. Set the cone angle to be about 83 and render the scene. The area of light
expands and is now too big.
13. Set the cone angle to be 50 and the penumbra angle of the light to be 10.
This will produce soft edges for the light.
14. In the Attributes menu, expand the Shadows section and further expand
the RayTrace Shadows Attribute section for this light. Turn on the Use
Ray Trace Shadows radio button. This will cause objects being lit by
this light to cast shadows when you render using ray tracing.
15. Set shadow rays and ray depth limit to be 3 to ensure that shadows are
created correctly.

16. Usually, shadows are never stark black. They are some shade of grey,
depending on the lighting. To achieve this effect, we also modified the
Shadow Color attribute to be a dark grey.
17. Create a similar spotlight for the shade on the other side. You will have
to position it appropriately.
18. Set this light's cone angle to be 60, penumbra angle to be 10 for a soft
light on the wall, intensity to be 0.8, and color to be a light yellow. You
should be able to see a cone of light on the wall when you render the
image.

Render the scene - wow! Doesn't it look amazing? And we still haven't gotten
to ray tracing. This will be the final step, when we will see reflections;
 refractions and shadows all come together to enhance the visual realism of our
image.

8.6 Finally - Ray Tracing

1. From the Shelf select the Display Render Globals Window icon.

This will bring up a window allowing you to set the rendering options.
2. Set the Render Using option to be Maya Software.
3. Under Resolution, set the resolution to 640 x 480 or higher if you wish.
This sets the rendered image to the specified size.
4. Click on the Maya software tab.
5 . Under the Anti-aliasing Quality options, set Quality to be Production
Quality-this will be a quality render.

6. Under Ray Tracing quality, set Ray tracing to be on!

Hurray. We are going to be ray tracing our scene!

You cannot use IPR to render ray-traced images.
7. Select Render>Render Current Frame from the menu bar.
You will see a ray-traced image being rendered of the currently selected
window. Be patient. Ray tracing can take a while. Preview your render. You may
need to modify your lights, models, and shaders to get the reflections and
shadows in the exact spots you want!
Only the spotlight for which we turned on the ray-trace shadows option, will
be obstructed by objects in its path, causing shadows. The reason we had this
light to be located outside of the shade is so that its light would not be blocked
 by the cylindrical cap! The other do not cast shadows, and hence, objects in their
path do not obstruct the beam of light. You can play around with the light settings
as you please. For the image in Color Plate 5, we actually turned on shadows for
the point light in lamp1 as well.
You can also pick specific models and edit their surface materials. For
example, we actually added some ambient color to the table and the glass so that
they didn't render as dark. Use the Window>Outliner to select objects as desired
if you have a hard time picking them otherwise. You can also find a Maya model
of the world that we modeled in a file called world.mb (Maya Binary file format)
under the installed directory for our software, in the folder Models.
Have fun experimenting. And yes, we did cheat and add in a texture mapped
cube in our color plate. You can see it reflected in the mirror. Define your own
texture image for this cube to see it in the final image.

Summary

In this chapter, we have learned how to use Maya to model and render a 3D
scene. Could you visualize what code was running in the background to display
the models and images? By now, you know enough to actually write a small
Maya package yourself!
The next section will deal with how to make our scene come to life by adding
animation. We shall also see how to add more pizzazz to our scene with
additional lighting and viewing techniques.
Section 3
Making Them Move
In the last few sections, we saw how to model objects and render them to create
photo-realistic images. Our next task is to give life to our creations by making
them move: animation!
When you see a cartoon on TV, you see characters in motion. This motion is
just an illusion: the strip itself is actually composed of a sequence of still images
(or frames) which when played back in rapid succession gives the illusion of
motion. The most popular way to create animation is by generating a number of
successive frames differing only slightly from one another. However, just
making the characters move is not enough to make an animation look believable.
Walt Disney himself put forward the seven principles of animation that add zest
and realism to the animation.
In Chapter 9, we shall look into the techniques and principles used to create
traditional (hand drawn) 2D computer animation and how they can be adapted in
a 3D world. Chapter 10 will walk you through the creation of an animated 3D
game. In Chapter 11, we shall use Maya and work through the process that
traditional production houses use to create our own animated production.

Hold on to your seats!

Chapter 9
Animation

To animate means to bring to life. In the last few chapters, we saw how we can
use the principles of Computer Graphics to draw and render 3D worlds. Our next
task is to breathe life into our models.
When you see a movie on television or in the theater, you see characters in
motion. This motion is usually smooth and continuous. The actual tape
containing the movie, however, consists of a sequence of images, or what are
known asfiames. These frames, when flashed in rapid succession on the screen,
cause the illusion of motion. Fig.9.1 shows such a sequence of frames in which
a smiley face changes into a frowning one.

Fig.9.1: A sequence of frames

The illusion of motion exists because of a peculiarity in the human vision
system. When you see a picture, it tends to remain in your mind's eye for a brief
period of time--a phenomenon called persistence of vision. If the pictures are
shown quickly enough, then a new picture arrives before the old one fades out.
Your eyes perceive a smooth transition between the frames. If the difference
between successive frames is too large or the frames are not shown quickly
enough, then the illusion shatters and the movement seems jerky. You may have
seen this kind of effect in older movies or documentaries. Sometimes the
QuickTime playback of streaming videos also displays this effect. Typically, a
film needs to display 24 frames a second (called frames per second or fps) to
look smooth. For video, 30fps is needed.
 As in movies, an illusion of motion in the CG world is created by playing
back a sequence of CG images. However, it is not enough to just make the
models move-their motion should look realistic. We shall present the basic
principles of animation that help achieve realism in motion. We begin our
animation journey by exploring the realm of traditional animation. We will then
see how these principles are carried over and applied to a 3D CG world as we
build our own animated movie.
In this chapter you will learn
w Traditional (2D, hand-drawn) animation
3D animation
W Applying the principles of animation to achieve realistic motion

9.1 Traditional Animation

In traditional 2D animation, most of the work is done by hand. The story to be
animated is written out on a storyboard. Initially only the key frames are
identified. Key frames are fiames selected by the master animators for their
importance in the animated sequence, either because of the composition of the
frames, an expression depicted by the characters, or an extreme movement in the
sequence.
Let us reconsider the smiley face sequence we saw in Fig.9.1. The key frames
of this sequence would be the frames with the smile and the final frown, as
shown in Fig.9.2.

Key Frame #I Key Frame #2

Fig.9.2: Key Frames in the smiley animation

Once the key frames are drawn, the next step is to draw the intermediate
frames. The number of intermediate frames, or in-betweens as they are called,
depends on the length of the animated sequence desired (keeping in mind the fps
count that needs to be achieved). This step of drawing in-betweens is also called
tweening.
If we were going to create a video of the face smiley animation, we would
need to create 30 fps. For a short five minute animation, we would need (30 fps)
x (60 seconds) x (5 minutes) = 9000 frames to be drawn! Now you know why
Disney hires so many animators to work on a single movie.
A technique that helps tremendously in the process of creating animations, is
called eel animation. When we create an animation using this method, each
character is drawn on a separate piece of transparent paper. A background is also
 drawn on a separate piece of opaque paper. Then, when it is time to shoot the
animation, the different characters are overlaid on top of the background. Each
frame can reuse the same elements and only modify the ones that change from
one frame to the next. This method also saves time since the artists do not have
to draw in entire frames-just the parts that need to change, such as individual
characters. Sometimes, even separate parts of a character's body are placed on
separate pieces of transparency paper.
Traditional animation is still a very time-consuming and labor-intensive
process. Additionally, once the frames are drawn, changing any parts of the story
requires a complete reworking of the drawings.
How can we make animation less laborious and more flexible? The answer
is found in the use of computers.
Computers are used by traditional animators to help generate in-betweens.
The animators sketch the key frames, which are then scanned into the computer
as line strokes. Many software programs are also available that enable drawing
directly on the computer. The computer uses the line strokes of the key frames
to calculate the in-betweens. Using sophisticated algorithms and equations, the
in-betweening can be made to look very natural and smooth.
The use of computers is not, however, limited to this minor role. Since the
advent of 3D Computer Graphics, computers are used in a big way to generate
the entire animation sequence in a simulated 3D world. The sequence can be a
stand-alone film, like Toy Story or The Incredibles, or can be composited into a
real action film, as in Lord of the Rings or Terminator-2.

9.2 3 0 Computer Animation - Interpolations

The most popular technique used in 3D computer animation is the key frame in-
betweening technique that was borrowed from traditional animation. In this
technique, properties of a model (such as its position, color, etc.) are identified
by a user (usually an animator) in two or more key frames. The computer then
calculates these properties of the model in the in-between frames in order to
create smooth motion on playback.
Computing the in-between frames is done by taking the information in the
key frames and averaging it in some way. This calculation is called interpolation.
Interpolation creates a sequence of in-between property values for the in-
between frames. The type of interpolation used depends on the kind of final
motion desired.
Interpolation can be used to calculate different properties of objects in the in-
between frames-properties such as the position of objects in space, its size,
orientation, color etc.

Let us learn some more about some basic interpolation methods and how we
can use them to define an animation.
Linear Interpolation
The simplest type of interpolation is linear interpolation. In linear interpolation,
the values of a given property is estimated between two known values on a
straight-line basis (hence the term linear). In simpler terms, linear interpolation
takes the sum of the property values in the two key frames and divides by the
number of frames needed to provide as many equally spaced in-between frames
as needed. For example, consider a model P, with a position along the y-axis of
Yi, at the initial key frame Kin.At the final key frame, Kfin,they position has
moved to Yfin.Now, say we want only one in-between frame f. This frame would
be midway between the two key frames, and P would be located midway
between the positions Yi, and Yfinas shown in Fig.9.3.

Fig.9.3: Linearly interpolating the Y position

Mathematically, the position Y(f) at Frame f, is given by

YO=Yin + (Ylin - Yi,J / 2 = + YiJ / 2
Now, suppose we want two in-between frames, fl and f2. The model would
then be defined such that at fly it occupies a position, Y(fl), which is one third
the distance from Yin; and at frame f2, it occupies a position Y(f2), which is two
thirds the distance from Yin.

?
Fig.9.4: Two inbetween frames
 Mathematically,
Yfl) =Yin + (YJin - YiJ / 3

In total, we would have four frames: Kin, fl, f2, and Kfin-enough to make
an animation that lasts for 116th of a second (assuming 24 frames per second).
Animations are usually represented in terms of timelines. The animation
timeline defines the frames of the animation along an axis. The property values
for the active object at these frames can be depicted along this axis either
visually or just as values. For our example, we could define the following
animation timelines:

Kin

Kin
Fig.9.5: Animation timeline
If we were to extend our discussion to a total of N frames (including the
initial and final key frames), we would have N-2 in-between frames. The
position Y at any frameJ can be can be calculated by the equation

05fl.N-I
At f r a m e p 0, we attain Y(0) = Yin: the initial key frame.
At frame f = N- 1, Y(N- 1) = Yfin: the final key frame.
The N-2 in-between frames can be calculated by varying f from 1 to (N-2).
As we vary f , we linearly interpolate the position of P from Yin to Yfin.
Do you recognize this equation? Yes-it's the parametric equation of a line!
The parameter we use is the frame number.
We can define similar equations for interpolating the position of P along the
x- and z-axis:
x f l =xin + f " ( X j , - xi,)/ (N - I)
zfl=zin +f"(Zjn - ZiJ (N- 1)
OLf(N-I
 This set of equation enables us to linearly interpolate the position of model P
between two key frames. We can apply the same equations to interpolate rotation,
scale, color, etc of the object. What we are doing is calculating the transformation
at each in-between frame using a linear interpolation equation. The code to
calculate the intermediate value is as shown below and can be found in the file
linearinterpolation.cpp,under the installed directory for our sample code:

GLfloat InterpolatedValue(GLfloatKFinitvalue, GLfloat KFfinalvalue, int N,

int f){
GLfloat ivalue;
ivalue = KFinitvalue + fC(KFfinalValue-KFinitValue)l(N-I);
1
In the above equations, we assume that the first frame of the sequence is frame
0. However, this may not always be so, especially if we have more than two key
frames. A more general version of the interpolation equation, assuming a total of
N frames and starting with an initial frame of InitF is as follows:
X@ =Xin + Cf- Initt) * (Xfi - Xi,, / (N - 1)
Y@ =Yin + Cf-Initt) * (Yjn- Yi,J / (N - I)
Z@ =Zin + Cf-Initt) * (Zfin - Zin) / (N - I)
I n i e f,< Initf + N - I

This equation can be calculated by our function InterpolatedValue, if we pass in

the parameter f as the value, (currentframe-InitJ)

Example Time
Let us take the example of a bouncing ball to illustrate how linear interpolation
works. We will initially move the ball only along the y-axis to grasp the
concepts. Then we will expand the motion along the x-axis as well.

LJ A And C

Timeline
Fig.9.6 : Key frames and graph of a bouncing ball
 As shown in Fig.9.6, a ball is initially at position A=(O,Ya,O). It goes up to
position B=(O, Yb,O) and the falls back to Position C=(O, Ya,O). The motion of
the ball can be drawn as a graph of its position in Y against the animation
timeline, also shown in Fig.9.6.
The position of the ball at A, B and C define our key frames. Let us say, we
want our animation to be 20 frames long. Further, let the three key frames be at
frames 1,10, and 20 along the animation timeline. To define the linear
interpolation between positions A and B, we go through 10 frames from 1 to 10.
The interpolated values of Y for any frame f is defined as
Yfl = Y,+f-I) *(Yb-YJ(I0-I)
1sf 5 1 0
In this same sequence, positions for frames between 10 and 20 need to be linearly
interpolated between positions B and C. These positions can be expressed as
Yfl = Yb+f-l 0) *&- Yd/9
10< f 1 2 0
Using the two equations, we are able to define the position of the ball at any
intermediate h m e in the animation.To generalize the code to calculate these equations,
assume we stored the positions of the key fi-ames in an array called KeyFramePositions[].
In our example, KeyFramePositions = {0,10,20}.We store the property values in an array
called KeyFrameValues[], which in this case = {Ya, Yb, Y, }. Then we can define a
h c t i o n that can calculate the intermediate value at any given h e as

GLfloat EvaluateLinearAt(GLfloat *KeyFrameValues, GLint *KeyFramePositions, GLint

NumOfKeyFrames, GLint fl
{
int i:
lint N, InitF;
GLfloat value;

I1 Find the two keyframes which this frame interpolates

for (i -0;icNumOfKeyFrames;i + +I{
if If c KeyFramePositions[ill
break;
?I--;
if (i<OI retum (KeyFrameValues[Oll; 11should not happen
11 Define N and lnitF

--
N KeyFramePositions[i + 11-KeyFramePositions[il;
InitF KeyFramePositions[il;
11 interpolate between KeyFramePositions[il and KeyFramePositions[i+ 11
-
value InterpolatedValue(KeyFrameValues[i], KeyFrameValues[i+ I],N, If-InitFI);
retum value;
 The code can also be found in the file provided: 1inearinterpolation.cpp.
In Example9-1, we define a ball bouncing up and down. Its position at
KeyFrame A is Ya = 0; at B, Yb = 25; and at C, Yc = 0 again. We define

-
-
GLFloat "KeyFramePositions {I,I0,20).
Glfloat *KeyFrameValues {0,10,0)

- -
int NumberofKeyFrames 3;
int MAXFRAMES 20;

If we let the glut sphere represent our ball, then the code for this animation looks
as follows:

-
ypos EvaluateLinearAt(KeyFrameValues,KeyFramePositions,
NumberofKeyFrames, currentFme1;
glloadldentity (I;
gluL00kAt(0.,0.,13,0,0,-100,0.,1 .,O.I;
glTranslatef(O.,ypos,O.I;
glutWireSphere(.5,10,10I;

where currentFrame is updated every loop by a timer callback. The entire

code example code be found in Example9-1/Example9~1.cpp.
Let us make the ball bounce some more. Due to the nature of gravity, each
time the ball bounces back up, it bounces to a point lower than the prcvious
bounce. To accomplish this, we extend our animation to 70 frames, and add in a
six more key frames at D, E, F, G, H, and I as shown in Fig.9.7. The only changes

Fig.9.7: Key Frames for Bouncing Ball

you need to make to the Example9-1 is to define the kcy-frame arrays as:

-
GLfloat KeyFrameValues[91 {0,10,20,30,40,50,60,70);

-
int NumberofKeyFrames 9;
-
GLint KeyFramePositions[91 {0,10,0,7,0,4,0,1,0);

-
int MAXFRAMES 71;
Voila! You will see a ball bouncing.
Let us now add in some motion along the x-axis, so that the ball appears to have
been thrown by someone, and is slowly bouncing to rest.
In order to accomplish this, we define a set of key frames for motion along the
x-axis as

KeyFrameXPosValues[O]
KeyFrameXPosValues[ll
-- 0;
3.5;
KeyFrameXPosValues[2] - 6;
KeyFrarneXPosValues[3]
KeyFrameXPosValues[4]
-- 8;
9.5;
KeyFrameXPosValues[5] - 11;
KeyFrameXPosValues[61
KeyFrameXPosValues[7]
-- 12;
12.5;
KeyFrameXPosValues[8] - 13;

We now interpolate the ball's x position as well as they position by calling the
interpolation function

ypos - EvaluateLinearAt(KeyFrameValues, KeyFramePositions, NumberofKeyFrames,

currentframe);
xpos - EvaluateLinearAt(KeyFrameXPosValues, KeyFrarnePositions, NumberofKeyFrames,
currentFrame);
glloadldentii 0;
gluL00kAtl0.,0.,13,0,0,-100,0.,1.,0.);
glTranslatef(xpos,ypos,O.);
glutWireSphere(.5,10,10);

This example can be found in Example9-2/Example9-2.cpp.You will see the

animated ball moving along the x- and y-axis as defined.
In real life, we do not expect the ball to actually follow a straight-line
trajectory along its path. Also, at the top, where the ball reaches its highest point,
we expect the ball to slow down before it changes direction. The motion of our
ball is a bit jerky and abrupt.
The key drawback of linear interpolation is that it is based on constant speed
between key frames, and the interpolation defines straight-line paths. This makes
the animation looks jerky. Let us see how nonlinear interpolation can rescue us
from this predicament.
Let us see how non-linear interolation can rescue us from this predicament.

Non Linear Interpolation

For most animation situations, linear interpolation is not an effective animation
technique. More complex non-linear interpolation techniques are employed for
simulating effects in nature.
 Recall the nonlinear paramateric equations from Chapter 7.

As usehl as these equations are in modeling, they are even more usehl in
animation! Instead of interpolating vertex points of a model, we can interpolate
key frame positions of an animation using the very same equations. The
parametric cubic equation set can be replaced to use the frame, f, as the
interpolating parameter:
X f l = a d 3 + bJ2 + cJ+ dx
Yfl = a d 3 + bJ2 + cJ+ dx
Zfl = a d 3 + bJ2 + cJ+ dx
f = Cf-mi@/(N-1)
Initf_<f,<N=>O5f_<1

The coeficients define the type of cubic spline being used. We covered
bicubic splines in Chapter 7. Bicubic splines are great for modeling, but they can
be cumbersome to use, since they only approximate the two interior points. In
this chapter, we shall use coefficients that specify a derivative of a hermite cubic
spline called the Cardinal cubic spline. The advantage of this spline is that the
curve interpolates all the points specified. We need at least four key frames in
order to define a cubic interpolation. In the file, cubicinterpolation.cpp, we
define a function to evaluate a Hermite cubic interpolation

-
GLfloat tension 0.3;
GLfloat CardinalValue(GLf1oat KFValuernl, GLfloat KFValuel, GLfloat KFValue2 GLfloat KFValuep2, int
N, int f1
{
GLfloat F - 1.O*fl(N-11;
GLfloat tangent1 - I1.-tension1*(KFValue2-KFValuern1112.;
GLfloat tangent2 - I1.-tensionl*lKFValuep2-KFValuel112.;

GLfloat hvalue - F*F*F*(2.*KFValuel-2.*KFValue2+ tangent1 + tangent21 +

F*F*
(-3.*KFValuel+3.*KFValue2 -2.*tangentl - tangent21 +
F*(tangentlI+
KFValuel;
return hvalue;
 The tension variable controls the tangents at the control points. Smaller
values of tension tend to produce tighter curves-the motion at the control point
is more abrupt. Increasing- the value of tension will relax the curve at the control
points, leading to slower motion around the control points.
For the bouncing ball from Example9-2, if we applied hermite interpolation,
the animation plot would appear as follows:

Gngent vector

1 10 20 30 40 50 60 70
Time linelframes
Fig.9.8: Hermite Interpolation Timeline

You can see how the ball's motion slows down at the control points.
In cubicinterpolation.ccp, we also define a function:

Glfloat EvaluateCubicAt(G1float *KeyFrameValues, Glint

This function figures out which four control points to send in to the
Cardinalvalue function. It also appends a copy of the first and last points to our
keyFrameValues array to interpolate the first and last points correctly. In
Example9-3, we modifL our ball animation to now use the cubic interpolation:

ypos - EvaluateCubicAt(KeyFrameValues, KeyFramePositions,

NumberofKeyFrarnes,currentframe);
xpos - EvaluateCubicAt(KeyFrameXPosValues, KeyFramePositions, NumberofKeyFrames,
currentframe);
glloadldentity I);
gl~L00kAt(0.,0.,13,O,O,-100,0.,1.,0.~;
glTranslatef(xpos,ypos,O.);
glutWireSpherel.5,10,10);

Watch the ball bounce. Does the animation look smoother? We achieved the
desired slow-down of the ball at the top peaks. However, do we really want the
same effect at the bottom? The ball should actually speed up at the bottom! Try
changing the value of the tension parameter to see how the motion is affected at
the endpoints. Most animation software employs equations that let the user
interactively tweak the interpolation graph at every control point in order to
achieve desired behavior.
There are cases when cubic interpolation does not work too well. Consider a
situation where, in key frame D, we want the ball to stay on the ground as shown
in Fig.9.9. The graph shows a wiggle between points C, D and E that makes the
graph go below the 0 level we defined for the ball. If we had the ball animating

TimelineIFrames
Fig.9.9: A wiggle below the ground

on top of a ground plane, then the ball would go underneath it! We would need to
add in extra control points to avoid this behavior. Try this with your ball animation
by changing the KeyFrameValues array and see what happens. Try it for the linear
interpolation case as well. Unlike linear interpolation, cubic interpolation does not
retain the values of adjacent key frames when these are equal.
It is hard to achieve accurate realism using only one interpolation technique.
In most cases, a combination of different schemse is needed to get good results.
Extensive ongoing research constantly delivers better ways to make realistic
interpolations possible. In any production, the animator has a big role in
understanding how interpolation works and how to modify the interpolation
graphs to achieve desired affects.

Animating Snowy
Let us use our interpolation equations to create a slightly more complicated
animation. Recollect our snowman model from Chapter 5. We shall use it to
create an animation of a snowman bouncing on the ground.
In Example9-4, we reuse the snowman model from Chapter 5. We changed
the snowman model code to comment out transforms that we did not need - for
sake of speedier display. Again for speed considerations, we shall preview our
snowman animation in wire frame mode.
In this example, we will only animate the root level component of the
snowman - the null node. Recall that the null node is the parent of all the
components of the snowman. Its pivot point is at the base of the snowman.
Recall that this node (as well as each component of the snowman) had a
transform array applied to it. These transforms were specified in an array of size
9 storing the (Tx,Ty,Tz,Sx,Sy,Sz,Rx,Ry,Rz) values:

-
GLfloat snomanXforms[91 {0.,0.,0.,0.,0.,0.,1.,1.,1.);
-
GLfloat botXforms[9] {0.,0.,0.,0.,0.,0.,1.,1.,1.);

In this example, instead of using a one-dimensional array of the nine xform

values, we use a two dimensional array, to hold the 9 xform values for each key
frame we define for the snowman (we assume a max of 10 key frames) This
leads us to define the following structure to set the xform values for the
snowman at its key frames:

GLfloat smKeyFrameValues[l0][9]; 11 [frame][xform values]

The array to hold the key frame positions and number of key frames defined
for the snowman are defined as follows:

GLint smKeyFramePositions[101;
//This array holds the actual positions of the key frames defined for the snowman
GLint smNumKeyFrames;
//The number of key frames that the snowman has

We define a utility function to actually initialize the key frame values for the
snowman.

I/ Set the key frame at frame-kf with the specified transforms

void SetKeyFrameiint kf, GLfloat keyframevalues[l51[91, GLint
"keyframepositions, GLint "numKF,GLfloat Tx, GLfloat Ty, GLfloat Tz,
GLfloat Rx, GLfloat Ry, GLfloat Rz ,GLfloat Sx, GLfloat Sy, GLfloat Szl
{
keyframepositions["numK~ kf;
keyframevalues["numK0[0] Tx;
--
keyframevalues["numKA[1] Ty;
keyfrarnevalues["numKfl[2] Tz;
--
keyframevalues["numKfl[3] Rx;
keyfrarnevalues["numKfl[4] Ry;
--
keyfrarnevalues["numKfl[5] Rz; --
keyfrarnevalues["numKfl[6] Sx;
keyfrarnevalues["numKfl[7] Sy;
keyfrarnevalues["numKF1[8] Sz;
--
I/ increment number of keyframes defined for model
-
"nurnKF "numKF+ 1;
1
 To evaluate the position of the snowman at any intermediate frame, we define
a fiction

void EvaluateXformsAtlGLfloat KeyFrameValues[lO][9], GLint "KeyFramePositions, GLint

NumOfKeyFrames, GLint f, GLfloat *xformsl
{
11 interpolate between KeyFramePosiiions[il and
KeyFramePositions[i + 1I
--
N KeyFramePositionsF + 11-KeyFramePositions[il + I;
lnitF KeyFramePositions[i];

This function figures out the appropriate key frames to interpolate between
and then calls the linear interpolation function to evaluate the transform for the
current frame. This transform is then stored in the parameter array: xforms. Let us
bounce Snowy up and down on a snowy ground. We define our world with it's
extents being about -2 to 15 units along the x- and 0 to 12 unit along the y-axis.
We define a snowy ground plane as a solid cube that is grayish in color.

void draw-Ground(){
glPolygonModelGL-FRONT, GL-FILL);
glPolygonModelGL-BACK, GL-FILL);
glColor3f(0.8,0.8,0.8);
glPushMatrix0;
glTranslateflO.,-5.5J.l;
glScalef(25.,10.,10.);
glutSolidCubeI1.I;
glPopMatrix0;

glPolygonMode(GL FRONT, GL LINE);

g l ~ o l y g o n ~ o d eGL-GNEI;
l~~~~~~,
 The ground is based at y=O, but does not extend all the way to the end of the
world coordinates. We will define 45 frames in this animation sequence. We
want to open the sequence with Snowy jumping into the scene. To do this, we
define a key frame for Snowy at frame 0, with his Tx=-5 and Ty =6.

At frame 15, we want Snowy to hit the ground.

At frame 30, Snowy rises back up again, and at frame 45 he lands back down-
but oh no! This time he has not landed back on the ground!

A timer function periodically updates the current frame number and calls the
Display function. The Display function merely calls the EvaluateXformsAt to
evaluate the transforms of the snowman for the current frame and calls the
snowman drawing code with these transforms.

void Display(void)
{
GLfloat ypos, xpos;
glutSwapBuffers0;

EvaluateXformsAt(smKeyFrameValues, smKeyFramePosiiions, smNumOfKeyFrames,

currentframe, snomanXformsl;
draw SnowMan(snomanXforms, botXforms, turnxforms, headxforms, IEyeXforms, rEyeXforrns,
nosexfork, IHandXforrns, rHandXforms1;
 Remember the use of the Push and Pop matrix-this ensures that when we pop
back from the snowman drawing routine, we are back in the transformation state
that we started with. So we do not need to adjust the position of the ground.
We have the basic motion of Snowy down. If you run the program, you will
notice that Snowy doesn't look very natural just facing us while he is bouncing.
To make him look in the direction he is going add in a rotation about y of about
40 degrees. View your animation. We want to add more to this animation. We
would like Snowy to tilt back as he lands and reorient himself for takeoff. To do
this, we redefine the key frames as follows. At frame 0, we just tilt Snowy back
by rotating him by 20 degree around the Z axis.

At frame 10, he lands on the ground:

At frame 13, he straightens himself up:

At frame 16, he reorients himself for take off:

SetKeyFramell6,smKeyFrameValues, smKeyFramePositions, &smNumOfKeyFrames,

4.,0.,0., 0.,40.,-20,1.,1.,1.1~
At frames 26 and 30 Snowy reaches the highest point of his bounce and re-aligns his body:

In frame 40, Snowy lands (but not on the ground); we clamp his motion on
frame 45 so he won't move any farther for the rest of the animation:
 View your animation. You can add in surface properties to the snowman and
light up your scene to see a solid bouncing snowman.
The code can be found under Example9-4 in files, Example9-4.cpp7
Smowman.cpp and Sn0wman.h. The animation looks good; but it can still do
with some zing!

9.3 The Principles of Animation

Three-dimensional computer animation has made great advances in the last few
years. There are numerous animation packages (we shall look into Maya in the
next chapter) that can help even a novice develop quality animations. However,
it is not uncommon to see animations that lack zing-they seem drab, look unreal
and reek of CGism. Here is where the principles of traditional 2D animation
come to the rescue.
The concepts used in hand-drawn animations, to make the sequence look
believable, can be used in the 3D realm to add as much zest and depth into the
animation as we see in its 2D counterpart.
Walt Disney himself put forward many of these principles of animation
which is now used as a bible by animators worldwide. Refer to LASS87 for a
more detailed analysis of these principles.
In this section, we shall explore some of these principles and see how we can apply
them to Snowy, whom we animated earlier. We will progressively add more key h m e s
into our sequence from Example9-4, to enhance the appeal of the bouncing Snowman.
The (entire) modified code can be found under Example9-.5/Example9_5.~pp.

Squash and Stretch

In real life, only the most rigid objects retain their original shape during motion.
Most objects show considerable movement in their shapes during an action. If
you notice how a ball bounces on the ground, it seems to squash on impact and
stretch back up when rising. The amount of squash depends on how rigid the ball
is-a soft rubber ball squashes a lot more than a hard baseball.
One trick used in animation to depict the changes in shape that occur during
motion is called squash and stretch. The squashed position depicts a model form
flattened out by an external pressure or constricted by its own power. The
stretched position shows the same form in a very extended or elongated
condition. Squash and stretch effects also help to create illusions of speed and
rigidity.
Human facial animation also uses the concepts of squash and stretch
extensively to show the flexibility of the skin and muscle and also to show the
relationship between parts of the face. During the squash and stretch, the object
should always retain its volume. If an object squashes down, then its sides should
stretch out to maintain the volume.
 The standard example of this principle is the bouncing ball. Fig.9.9 shows
some of the frames in a bouncing ball example. The ball is squashed once it has
made impact with the ground. This gives the sense of the material of the ball. The
softer the ball, the greater the squash. The ball is elongated before and after it hits

Fig.9.10: Squash and Strecth applied to a bouncing ball

the ground to enhance the feeling of speed. Hence, squash and stretch effects
help to create illusions of speed and rigidity. Its one of the most important
principle used in animation to simulate realism. Human facial animation also
use the concepts of squash and stretch extensively to show the flexibility of the
skin and muscle and also to show the relationship between parts of the face. We
will use the principles of squash and stretch to add some life to our Snowy
animation.
We will be using the scaling transforms to add squash and stretch to Snowy.
At frame 10, Snowy is just about to land on the ground. Let us stretch him up at
this frame to enhance the feeling of him speeding up in motion before impact
with the ground. To do this, set Sy= 1.2. Since we wish to retain volume, we will
squash him down in x- by setting SX= 0.8:

SetKeyFrame(1O,smKeyFrameValues, smKeyFramePositions,
&smNumOfKeyFrames, 4.,0.,0., 0.,40.,20.,0.8,1.2,1.1;

At frame 13, Snowy has made maximum impact with the ground. He should be squashed to
- -
depict this force. Set Sx 1.2 and Sy 0.8

Finally, at frame 16, Snowy is ready to launch back into the air. Set his Sy =
1.2 and Sx = 0.8 to give the illusion of him stretching back up:

At frame 40, we will make Snowy stretch as if he is about to land on the ground
(even though he really isn't!)
 View your animation. Does the animation start looking more realistic
already? And we still have more to go!

Fig.9.11: snowy: Squashed and then ready ot bounce up again!

Staging
Another important concept in animation is staging. Staging is the presentation of
an idea so that it is completely and unmistakably clear. A personality is staged so
that it is recognizable; an expression or movement is brought into focus so that
it can be seen. It is important to focus the audience on only one idea at a time. If
a lot of action is happening on the screen at once, the eye does not know where
to look, and the main idea will be upstaged. The camera placement is very
important to make sure that the viewer's eye is led exactly to where it needs to
be at the right moment. Usually, the center of the screen is where the viewer's
eyes are focused. Lighting techniques also help in staging.
Let us apply staging to our Snowy animation. We shall make Snowy
complete his second bounce only to find himself off the plateau and standing on
thin air! To stage this part of the animation, we want to move the camera to zoom
in on Snowy's current location. Such a swing in the camera is called a cut.
A cut splits a sequence of animation into distinctive shots. The transition
from shot to shot can be a direct cut, which we shall use in this example. It could
also be a smooth transition, which is something we will look into in a later
chapter. Usually, the last few frames from the first shot are repeated in the second
shot to maintain continuity between the shots. Let us implement these concepts
to refine our Snowy animation. First, we increase the total number of frames in
our animation:

int MAXFRAMES - 61;

The cut occurs at frame 41, when Snowy just lands on thin air. We can
implement this cut by simply moving our camera position in the Display
function as follows:

if IcurrentFrame -- 41 I { 11 cut
glLoadldentity0;
gl~L00kAt(l6,5.,12, O,O,-100,0.,1 .,O.);
1
else if IcurrentFrarne -- 0) { I1 orginal location of camera
glLoadldentity0;
gluLookAt(8,7.,17, Or&-100,0.,1 .,O.);
1
How would you implement the camera repositioning in the case of a smooth
transition between the shots? Snowy himself needs more key frames defined. We had to
increase the size of all the key fi-ame arrays to accommodate for these new key fi-ames.

We also modified the parameter to the SetKeyFrame fimction appropriately.

Frames 41 onwards forms cut2 of the sequence. From frames 41 to 50, we repeat
the last bits of the animation from shot1 :

We clamp the animation from frames 50 to 60 for now:

You can go ahead and view your animation or wait until we add in more features.
(Yes, we agree: the cut looks jerky! We will show you how cuts are made smooth
in the next chapter)

Anticipation
Anticipation involves preparing the objects for the event that will occur in the
later frames of the animation sequence. If there is no anticipation of a scene, the
scene may look rather abrupt and unnatural. The anticipation principle makes the
viewer aware of what to expect in the coming scene and also makes the viewer
curious about the coming action. Anticipation is particularly necessary if the
event in a scene is going to occur very quickly and it is crucial that the viewer
grasp what is happening in the scene.
Continuing our Snowy animation, we already have the stage set up for the
audience: they have seen Snowy land on thin air. To create more anticipation, we
shall make Snowy look down to give an impression that he is trying to fig. out
what he is landing on. This would draw a viewer's attention to the fact that good
old Snowy is going to have a free fall from this frame onwards.
The key frames for the head from frames 1 to 45 need to be clamped to their
identity values:

From frames 47 to frame 55, Snowy looks down to see why he hasn't squashed:

Let us wait to finish our animation before finally viewing it.

Timing
Timing refers to the speed of an action. It gives meaning to the movement in an
animation. Proper timing is crucial to make ideas readable. Timing can define the
weight of an object: a heaver object is slower to pick up and lose speed than a
lighter one. Timing can also convey the emotional state of a character: a fast
move can convey the sense of shock, fear, apprehension, and nervousness, while
a slow move can convey lethargy or excess weight.
In the Snowy animation, we let Snowy look down lazily, using 8 frames to
look down (from frame 47 to frame 55). However, when he realizes his coming
plight (free fall) we want him to look up in shock. We shall make him look up in
only two frames to convey this sense. We also define his eyes to enlarge to
exaggerate the sense of shock:

SetKeyFramel55,sheadKeyFrameValues, sheadKeyFramePosiiions,
&sheadNumOfKeyFrames, 0.,0.,0.,40.,-20.,0.,1 .,I.,I.I;
SetKeyFrame(57,sheadKeyFrameValues, sheadKeyFramePosiiions,
&sheadNumOfKeyFrames, 0.,0.,0., 0.,-20.,0.,1 .,I.,I.I;
Almost there-one last trick and we will be able to view a very realistic animation.

Secondary Action
A secondary action is an action that results directly from another action.
Secondary actions are important in heightening interest and adding a realistic
complexity to the animation.
For Snowy, a secondary action would be his hands moving up and down as
he bounces up and down. This motion of his hands would help him maintain his
balance and also add some realism to his bouncing. His hands would move down
when he is landing and raise up as he bounces up. When Snowy finally falls, his
hands swing upwards to convey the sense of free fall that he is in. The key frames
to rotate the hand upwards as Snowy falls down are shown below. We leave the
rest of the secondary motion as an exercise for you.

Ill keyframes for the left hand

SetKeyFrame(62,slHandKeyFrameValues,
slHandKeyFramePosiiions, &slHandNumOfKeyFrames, 0.,0.,0.,1 .,I.,I.,0.,0.,0.1;
SetKeyFramel65,slHandKeyFrameValues,
slHandKeyFramePosiiions, &slHandNumOfKeyFrames, 0.,0.,0.,1 .,.,I
I .,0.,0.,-90.1;
SetKeyFrame(70rslHandKeyFrameValues,
slHandKeyFramePositions, &slHandNumOfKeyFrames, 0.,0.,0.,1 .,I.,I.,0.,0.,-90.1;

Make the animation and play it. The code can be founder under Example9-5,
in files Example9-5.cpp, Snowman.cpp and Sn0wman.h.
Watch the animation and take special note of the tricks you have used in this
section to make the animation more fun and realistic. Particularly notice the
change in camera viewpoint, the roll of Snowy's head, his eyes widening, and his
hand movements, too. Interested readers should also try changing the
interpolation technique for Snowy to be non-linear. What do you see? Is the
motion better or worse than its linear counterpart? Add shading and lighting to
your animation as well for a cool animation.
There are several other animation principles that can be applied to achieve
realistic motion. Exaggeration or deliberate amplification of an idealaction is
often employed to focus the viewer's attention. Another common technique used
for achieving a subtle touch of timing and movement is called slow in and slow
out. In this the object slows down before reaching a key frame. We saw how this
makes for a natural look of the bouncing ball when it reaches its highest point.
The use of all these animation principles, along with good presentation of a
theme, can help produce eye-catching and entertaining animation.
9.4 Advanced Animation Techniques
Interpolation is just one of the techniques used to generate realistic animations.
Depending on the effects desired, different kinds of animations can be combined
to achieve desired effects. We discuss some of the more advanced animation
techniques briefly in this section.

Dynamics
Dynamics is the branch of physics that describes how an object's physical properties
(such as its weight, size, etc) are affected by forces in the world that it lives in (like
gravity, air resistance, etc). To calculate the animated path, an object traversing the
virtual 3D world is imbued with forces that model after the physics of the real
world. These forces act on the models causing some action to occur. For example,
blades of grass sway in the wind depending on the speed of the wind, and to a lesser
extent, depending on the size of the blade.
In the 3D world, a wind force can be defined which when applied to models of
grass, affects their transformation matrix in a manner defined by the laws of
physics.. This kind of simulation was used in A Bug's Life, where an infinite number
of grass blades were made to constantly sway in the background by defining wind
forces on them. The foreground animation of Flick and his troop was still defined
using Key Frame animation.

Procedural motion
In this technique, a set of rules is defined to govern the animation of the scene.
Particle systems are a very good example of such a technique. Fuzzy objects like
fire and dust or even water can be represented by a set of particles. These particles
are randomly generated at a defined location in space, follows a certain motion
trajectory, and then finally die at different times. The rules define the position,
motion, color and size of the individual particle. A collection of hundreds of

Fig.9.12: Particles representing a water fall

240 9 4 ADVANCED
ANIMATION TECHNIQUES

particles can lead to realistic effects of fire, explosions, waterfalls, etc.

Motion Capture
Real-time motion capture is an advanced technique that allows animators to
capture real-live motion with the aid of a machine. An actor is strapped with
electronic nodes that transmit signals to the computer when the actor moves. The
computer records this motion. The captured motion data can then be applied to
the transformation matrix of the computer-generated characters to achieve a real
-live effect.

Kinematics
Kinematics is the study of motion independent of the underlying forces that
produce the motion. It is an important technique used to define articulated fig.
animations.
Usually, a defined model is applied that has a skeleton structure consisting of
a hierarchy of rigid links (or bones) connected at joints. Each link corresponds
to certain patches of the model surface, also called the model skin.A joint can be
rotated, causing the corresponding surface and the attached children bones to
rotate about it. There are usually some constraints on how much each joint can
be rotated.
This motion corresponds closely with the human anatomy, and hence is used
very often for human like animations.
Consider the model of a leg, with a defined skeleton consting of three links
L1, L2, and L3 at joints J1, J2 and J3 as shown in Fig.9.13. J1 is the head of the
hierarchy. A rotation about J1 causes Ll-J2-L2-J3-L3 (and the associated skin)
to rotate as shown in Fig.9.14. You can also rotate about joints 52 or 53 to achieve
a desired position.
Can you see the relation between this organization of joints and links and the
human (or any other animals) leg? Can you visualize how this technique could

LS

Fig.9.13: Rotation of the skelecton joints and links

 be applied to our android? If you recall the example of the marching android, a
similar principle was used to rotate his legs. He just didn't have any skin.
In forward kinematics, the motion of all the joints are specified explicitly by
the animators. As the joints are rotated, the motions of the links in the chain are
determined indirectly.
In inverse kinematics, the animator only specifies the final position of the
links at the end of the hierarchy. Math equations are then used to determine the
orientation of all the joints in the path that lead to this link, with the rotation
constraints applied against each joint.

Summary
In this chapter, we have looked at a few of the common animation techniques
used to bring 3D models to life. We have seen the classic example of a bouncing
ball and how different interpolation techniques yield different animations. We
looked at some advanced animation techniques that can make the life of an
animator simpler.
3D animation can be very drab if we do not apply the principles of traditional
animation. We have studied these principles and used them to animate Snowy,
our friendly snowman. In the next chapter, we will see how to animate the
camera to produce a stunning game. In Chapter 11, we will create a small movie
using Maya.
Chapter 10
Viewpoint Animation

In the last chapter, we learned some tricks to make our animations look more
realistic and appealing. In all the previous examples, the position of the camera
or our viewpoint was fixed. We only did a one-time change to the camera
position: a cut in the animation sequence. Camera movement is not limited to
changing across shots only. A technique often employed in animation is to move
or transform the camera itself through a 3D world. This technique is called
viewpoint or camera animation.
Why would one want to animate the camera? Well, camera movement adds
valuable depth to the scene. Moving the camera through space causes objects to
move in different ways depending on their distance from the camera, cluing us
in to how they are spatially arranged in our 3D world space.
In games, the camera represents the player's eye: he can change his or her
viewpoint by moving the joystick or the keyboard. A lot of planning is required
to move the camera in a game. The speed, field of vision, and orientation of the
camera are critical because what the camera sees is what an observer would
perceive. The faster the camera moves the faster the observer feels he or she is
moving. The sense of speed due to change in position and orientation obtained
by camera motion can produce eye-catching and exciting effects. These
techniques are commonly used in games and motion ride simulators, where you
feel like you are moving through a real-life scene. Other spectacular examples of
viewpoint animation can be found in animations of the solar system, the motion
of atomic particles in a virtual model, etc.
What you will learn in this chapter:
How to move the camera through space
How to capture keyboard input to change camera positions
w How to implement your own 3D game complete with camera motion
and fog
244 10 1 ANIMATING
THE CAMERA
IN THE SNOWYANIMATION

Let us begin this chapter by seeing how to move the camera in our snowy
animation.

7 Animating the Camera in the Snowy Animation

The cameralviewpoint defines how the 3D scene is being viewed. The position
of the camera and its orientation defines the final image rendered onto the screen.
The position and orientation of the camera can be captured at certain key frames.
To move the camera, we interpolate these camera parameters for the intermediate
frames in exactly the same way we used interpolation for object transforms. For
each in-between frame, the camera transformation matrix is computed and then
used (in conjunction with the object transformation matrix) to compute the final
scene orientation.
Let us see how to animate the camera with the snowman example from the
previous chapter.
You can use two techniques to set the camera position: one is by using the
call to

and the other is by making calls to the gl transformation functions @Translate,

glScale etc) but in the opposite direction of the intended motion. Recall from
Chapter 5, that moving the camera is the same transformation as moving the
objects in the opposite direction. In OpenGL, the camera viewing matrix and the
model transformation matrix are combined into one for this reason.
We shall see how to animate the camera using both techniques. Which
technique you use is depends on how easy it is for you to think of your 3D world
and how to move within it. Camera transforms should be the last transformations
to be applied to the matrix stack, since they are applied to all the objects. This
means that the camera transforms should be issued first. Recall how we cut
between the two shots in the snowman animation in Example9-5:

if IcurrentFrame -- 411 { 11 cut

glLoadldentity0;
gluLookAt(l6,5.,12,0,0,-100,0.,1.,0.~;
1
else if IcurrentFrame -- 0) {"I
glLoadldentity0;
gl~L00kAt(8,7.,15, O,O,-100,0.,1.,0.~;
1
In Example10-1, instead of cutting between the two shots, we shall smoothly
pan the camera across, starting at a viewpoint position of (8,7,15) at frame 30,
and ending at the final position of (l6,5,12) at frame 50. We do not animate the
orientation (the look at point of (0,0,-100) and the normal vector of (0,1,0)) in
this example. For this example, we define a set of key frames to hold the camera
position from frames 0 to 30 at the initial camera position:

For the next 20 frames, the camera moves smoothly to its new position:

To prevent any drift between frames 50 and 70, we define an extra key frame
at frame 70:

Finally, in the Display code, we interpolate the camera position and set the value
accordingly. Note that we only interpolate the position of the camera.

Voila! Watch your snowman animation, with a smooth camera pan just before
Snowy falls from the cliff. Does it help the animation in any way? Already you
get a sense of perspective between Snowy and the plateau on which he is
jumping. If there were more objects in the scene, the sense of perspective
between the various objects would be even keener. The entire code can be found
under Examplel 0-I/ExamplelO-1 .cpp.
In ExamplelO-2, we use the glTransform routines to simulate camera
motion. We define the key frames as before, but we actually will be using the
rotation to simulate camera rotation as well. From frames 0 to 30, the camera is
at the same position as in Examplel 0-1.
246 10 2 BUILDING
UP A HEAL rlME 3D GAME

From frames 30 to 50, we move the camera and rotate it slightly along the x- and
y-axis to stage Snowy in the shot:

Finally, in the Display code, we interpolate the camera position values.

Then we apply the camera transforms using the gl transformation routines.

Notice that we apply rotations first, since we want the camera to move about its
local pivot point. The translations are applied in the opposite direction as
specified:

Watch the camera zoom in on Snowy. In general, this technique is easier for
animating the camera, and we shall use it more in the next few sections.

7 0.2 Building up a real time 3 0 game

The snowman animation was a simple example of a camera animation.
Animating the camera can attain even more interesting results. In games and
motion rides, viewers are given the illusion of moving around in the 3D world
from the perspective of their own eyes-that is, as if they are sitting on the
camera and moving it along with them. Let us explore how we can achieve this
result by building up a 3D game.
In ExamplelO-3, we develop a game that entails a user exploring a 3D
terrain. The user can move himself back and forth and reorient himself in this
terrain. The user will be allowed to control the motion of the viewpoint by
pressing on the arrow keys on the keyboard. (In a typical game setting, this
fimctionality would be provided by a joystick.) In this example, we use the
keyboard up-down keys to move the camera along the x-z plane and the left-right
keys to enable the user to re-orient himself left and right.
We shall not be going into the mechanics of how we designed this game since
detailed those points in Chapter 4. In general, you should follow that process
when designing your own games.

Camera Definition
First, let us see how to define our camera. In the last example we used a nine-
element array to hold the camera position. In this example, it will be easier to
think of the camera in terms of its position and orientation exclusively: it is rare
for someone to scale a camera in the middle of a game. We define a structure for
the camera as follows:

typedef struct -tCamera

{
GLfloat position[31;
GLfloat orientation[31;
) CAMERAINFO;
We then define an actual camera:

CAMERAINFO camera;

We define the initial camera position to be at a default position

void InitCameraPosition(CAMERAlNF0*camera){
--
camera-> posiiion[O] 0;
camera->position[ll 0.2;
-
camera->position[2] 0 ;

that is, slightly above the ground and looking down the z-axis.
To apply motion to the camera, we simply apply the gl Transforms we saw
earlier, using the current camera position:

void MoveCameraPosition(CAMERAlNF0camera){
glRotateflcamera.orientation[21,0.,0.,1 .I;
glRotatef(camera.orientation[ll, 0.,1.,0.1;
glRotatef(camera.orientation[0], 1.,0.,0.);
glTranslatef(-camera.position[O],-camera.poiion[l], -camera.position[211;
1
The main question is this: How do we update our camera position based on user
input in real time?
248 10 2 BUILDING
UP A REAL TIME 3D GAME

Camera Update
To update the camera position, we make use of the popular physics concept:

Distance traveled - velocii*elapsed time

Therefore:

Final Position - Initial Position + velocity*lelapsed time)

The camera position is updated at every tick of the game logic. The velocity of
the motion is determined at every tick by checking the state of the arrow keys on
the keyboard. If the up arrow is being pressed, then velocity is positive, causing
forward motion. If the down key is being pressed, velocity is negative, causing
backward motion. If nothing is being pressed, then velocity is 0, causing no
motion to happen. The difference between the last timer tick and the current one
gives the elapsed time.
The keyboard call-back function that we used earlier only kicks in when a state
change occurs - a key is pressed or released. The new formula we are needs us to
identify the state of the key at every tick. We make use of the Windows function:

SHORT GetAsyncKeyStatelintvKeyl

This function checks the state of the key defined by the parameter vKey-is
it being pressed or not. Unix platforms provide a similar library call.
For simplicity, we define our own Boolean macro:

This function tests whether the key identified the specified vkcode is currently
being pressed down and returns 1 if it is. For the up, down, left and right arrow
keys, the vkcode ID is defined as: VK-UP and VK-DOWN, VK-LEFT and
VK-RIGHT. Using this function, we can set up a test to determine key state and
identify the velocity appropriately. Note that only one key is assumed to being
pressed at a given tick.

if lKEYDOWN(VK-UP))
-
velocity 0.4f;
if (KEYDOWN(VK-DOWN))
-
velocity -0.4f;

Similarly, the orientation of the camera can be determined by using angular

velocity. We use the math equation:

Final orientation of the camera = initial orientation + avelocity* elapsed time

The left and right arrow keys define the angluar velocity as follows

if (KEYDOWN(VK-LEFT))
-
avelocii -7;
if IKEYDOWNIVK-RIGHT11
-
avelocity 7;

High resolution timing is supported in Win32 by the QueryPerformanceCounterO

and QueryPerformanceFrequency() API methods. The first call,
QueryPerformanceCounterO, queries the actual value of the high-resolution
performance counter at any point. The second hnction, QueryPerformanceFrequencyO,
will return the number of counts per second that the high-resolution counter
performs. To retrieve the elapsed time from the previous tick, you have to get the
actual value of the high-resolution performance counter immediately before and
immediately after the section of code to be timed. The difference of these values
would indicate the counts that elapsed while the code executed. The elapsed time can
be computed by dividing this difference by the number of counts per second that
the high-resolution counter performs (the frequency of the high-resolution timer).

In this example we are driving the viewer along a terrain. The viewer cannot lift
himself off the ground (motion along the y-axis). He also can only rotate his
viewpoint about the y-axis (along the x-z plane, called a roll), but not along the
x-axis (called a pitch) or z-axis (called a yaw). If you want to simulate a flying
motion, you will have to enable this motion as well.
The distance traveled along the x- and or z-axis as a hnction of the total
distance traveled depends on the orientation of the camera position as shown in
the Fig. 10.1.

Fig.lO.1: Distance traveled in the XZ plane

Using this, we can define the function to update the camera position:

void UpdateCameraPosiiion(CAMERAlNF0"camera1 {
 LARGE-INTEGER current'lime;
GLfloat elapsedTime;
GLfloat velocity- 0;
GLfloat avelocity- 0;
-
Glfloat d 0;

I1 if up arrow has been pressed, velocity is now foward

if (KEYDOWNIVK-UP11
-
velocity 0.4f;
if (KEYDOWNWK-DOWN11
-
velocity -0.4f;
if (KEYDOWN(VK-LEFT))
-
avelocity -7;
if (KEYDOWNIVK-RIGHT))
-
avelocity 7;

--
camera->orientation[Ol 0.;
camera->orientation[ll + avelocii*elapsedTime;

-
d elapsedTime * velocii;
-
camera->orientation[21 0.0;

--
camera->position[O] + d * ((GLfloat1sin(TORADIANS(camera->orientation[1]111;
camera->position[ll 0.2;
camera->position[21-- d * ((GLfloat)(coslTORADIANS(camera->orientation[ll111;
1
Note that the math functions sin and cos require the angles to bc in radians,
hence the conversion. A timer thread is used to call the Display function at every
tick. The function first updates the camera position, and then draws the world
from this new vantage point.

void Display(void1

glClear(GL COLOR-BUFFER-BIT I GL-DEPTH-BUFFER-BIT 1;

gl~oadlde&yO;
1
void timerlint value)
{
11 Force a redisplay ..,also contains game logic
glutPostRedisplay0;
11 Restart the timer
glutlmerFuncll0, timer, 01;
1

Defining the World

Let us now see how we can define a world to view with our camera! For this
viewpoint animation, we will make use of a textured ground as the land strip that
we are driving through.

The Ground
The extents of the ground plane extend from -100 to 100, and it is defined as
follows:

void draw-GroundO
{

So actually, we should clamp our camera motion to never go beyond the

maximum boundaries in x and z of (-100,100).
In keeping with our outer-space theme, we assign the ground a texture of a
bumpy surface, as shown in Fig.lO.2. This image is actually an image of the
Mars terrain that we downloaded from the Internet. You should be able to find
tons of images on the Internet for your projects. Just remember that texture
images needs to be a BMP file with a size that is a power of 2 such as 256, 5 12,
or 1024 etc. You may notice that we did not use the glTexCoord function to
define the mapping of the texture coordinates. Instead, we use OpenGL's
functionality to automatically generate texture coordinates.

Fig.lO.2: A mars terrain

Automatic Texture Generation

OpenGL has three automatic texture generation mode. One is used when doing
spherical environment mapping. The other two are eye linear and object linear.
In eye linear mode, a texture coordinate is generated from a vertex position by
computing the distance of the vertex from a plane specified in eye coordinates.
In object linear mode, a texture coordinate is generated from a vertex position by
computing the distance of the vertex from a plane specified in object coordinates.
We shall use object linear in this example.

We need to do two things:

tell OpenGL that we will use the automatic texture generation with object
linear. This is done by making a call to the function glTexGeni0 with the
parameter, GL-OBJECT-LINEAR
tell OpenGL which plane we will use.
How to get the plane equation is tricky. Every plane is specified by four
parameters and defined by the equation: P1*x+ P2*y+ P3*z + P4 = 0. Using
GL-OBJECT-LINEAR, the texture coordinate at an vertex (xO,yO,zO,wO) is
given by P1*xO+ P2*yO+ P3*z0 + P4.*w0 (wO is I usually.)
The P 1, P2, P3 and P4 values are supplied as arguments to glTexGen*() with
pname set to GL-OBJECT-PLANE. With PI, P2, P3 and P4 correctly
normalized, this function gives the distance from the vertex to a plane. For
example, if PI = P2 = P4 = 0 and P3 = 1, the function gives the distance between
the vertex and the plane z = 0.
The distance is positive on one side of the plane, negative on the other, and
zero if the vertex lies on the plane. Since our ground lies in the x-z plane, we
want to generate s and t coordinates by the vertex distance from the z=0 plane
and the x=O plane.
The code to automatically generate OpenGL texture coordinates can be
defined as:
254 10.2 B u l ~ v l n lUP
~A REAL TIME 30 GAME

Some Clutter
We randomly throw in a few cylinders to clutter up the scene. A cylindrical
model can be created by using the glu library command

glucylinder 1 GLUquadric* quad, GLdouble base, GLdouble top ,GLdouble height, GLint slices, GLint
stacks I;

quad: specifies the quadrics object

base: specifies the radius of the cylinder at z=0.
top: specifies the radius of the cylinder at z=height.
height: specifies the height of the cylinder.
slices: specifies the number of subdivisions around the z axis.
stacks: specifies the number of subdivisions along the z axis.

The quadric object need only be defined once using the function

In our game, the cylinders are drawn with a texture map wrapped around them,
as shown in Fig.lO.3. This map causes the cylinders to look like buildings with
lit-up windows

Fig.10.3: Texture map for the cylinders

void drawwCylinder(GLfloat *position){

 g~~ranslatef(position[0l,position[ll+3,position[21);
glRotatefl9O,l10,01;
gluCylinder( cylinder, 1.,1., 7., 30,30 1;

You can make the game a lot more interesting by defining more interesting
and complicated models in the 3D world.

Putting the World Together

We define a simple structure to store the position and type of all the models in
the scene.

typedef struct -tModels

I1
GLfloat position[3];
GLint type;
) MODELINFO;

The two types of models we define for our game are

Wefine TANOY 1
Wefine TCYLINDER 2

The models are stored in an STL vector map of thc MODELINFO struct.
vectorcMODELINF0 *> Models;

We now define a function called Initworld, which randomly positions Andy

within the extents of our world. It also creates the quadric object and the 25
cylinders in the world. Thc object of the game is to find Andy hidden amongst
all this clutter.

void InitWorldO{
int i;
SYSTEMTIME systime;
GetSystemTmeN?tsystime);
11 Seed random number generator
srandIsystime.~Minute*60+ systime.wSecond);
 noofshapes - ReadVRML("..\\ModeIs\\robot.wrl", &coords[Ol[Ol,
&normals[01[0],&indices[0][0],&nindices[0][0], &(noofpoly[Ol), MAXSHAPES, MAXCOORDS);

MODELINFO "Andy, *Cylinder[25];

-
Andy (MODELINFO *)malloc (sizeof(MODELINF0));
-
Andy->posiition[O] RandomPosO;
--
Andy->position[Ol 0;
Andy->position[l] 0;
-
Andy->position[ZI RandomPosO;

- -
Andy->position[Zl 0;
Andy->type TANDY;

Models.push backIAndy1;
-
cylinder g l ~ ~ e w ~ u a d 1;r i c l I* Allocate quadric object *I
gluQuadricDrawStyle( cylinder, GLU-FILL 1; I* Render it as solid *I
for (i-O;i<25;i+ +I{
-
Cylindeail IMODELINFO *)malloc (sizeofiMODELINF0));
-
Cylindeail-> position[Ol RandomPosO;
--
Cylindeail->position[ll 0;
Cylinder[#> position[ZI RandomPosO;
-
Cylinder[il->type TCYLINDER;
Models.push-back(Cylinder[il);
1
The code for the drawworld fbnction is shown below. It simply draws the
ground, and then loops through all thc models dcfined in the Models vector,
calling the appropriate drawing routines.

void drawWorldO{
vector<MODELINFO *>::iterator it;
gITexEnvfIGL-TEXTURE-ENV, GL-TEXTURE-ENV-MODE, GL-DECAL);

- ---
for (it Models.begin0; it! Models.end0; + +it){
if ((*it)->type TANDY)
draw-Andyll*it)->position);
else if (I*itl->type --
TCYLlNDERl
draw-Cylinder(l*it)->position);
1

The entire code can be found under ExamplelO-3, in filcs:

Example 10-3.cpp, Example 10-3.h, Models.h and Models.cpp. You will also
need to include the provided files, bmp cpp, bmp.h, vrmLcpp and vrm1.h to
compile and link this project.
Run the program, and travel through your world. Can you find Andy? You
will find that you can actually travel through the objects! In an actual game, we
would clamp the camera's motion so that we would not be able to do so.
Try adding more cylinders and other models to your program. As you add
more and more objects, you may also notice a serious slowdown in speed. This
is because there are so many objects in our scene! One serious drawback in
OpenGL is that OpenGL does not understand the notion of objects; it just knows
about the polygons that make up the models. So it actually tests all the polygons
of every model to check whether they should be drawn or clipped. Each cylinder
model itself has about 100 polygons. Making 100 calls per cylinder can degrade
performance a lot. Instead, if we could perform a simple check and not draw
objects (cull objects) that are out of view, we could improve performance
substantially.

Object Culling
Remember that our camera is defined as a frustum with a clipping range of about
50 units:

The easiest way to check whether (entire) objects are within the camera view is
to approximate the viewing range with a sphere. We can define the sphere that
encompasses the viewing frustum as a sphere located at the approximate center
of this frustum

with a radius of 50 units. Any object that is not within the boundaries of this
sphere will be eliminated. Recall from math that the distance of point (x,z) from
a center (cx,cy) can be computed as

If this distance is less than the radius of our sphere, then we are within the
viewing distance (or close to it); otherwise we are outside. We can use this test
to determine whether the object is positioned within this sphere or not. The code

boo1 WithinViewingSphere(GLf1oat "sphcenter, GLfloat *pos){

if I I(sphCentertO1-pos[O])*IsphCenter[O]-pos[OI)
+ (sphCentert21 - pos[2])*lsphCente~2]-pos[2])) < (50.*50.))
retum TRUE;
258 I f ) 2 BUILDING
UP A HEAL. TIME 3n GAME

return FALSE;
1
returns TRUE if an object located at a position defined by the arraypos is inside
the viewing sphere.
The drawworld code is now modified to test if objects are within the viewing
sphere. If not, we do not make the openGL call to draw them, effectively culling
them from the scene.

for (it - Models.begin0;if (!IWihinViewingSphere(sphCenter,

it! - Models.end0; +it) {
+
("it)->position))l
continue;
else {
/I draw the objects

Far clipping plane =50.f-

Near clipping plane =0.5

Fig.lO.4: Sphere approximating the viewing frustum

The structure holding the models is often referred to as the scene graph.
Usually the graph is structured as a tree. Real objects occupying a certain region
of space in the scene are placed at leaf nodes. Nearby children are grouped
together under a parent node. The parent node (usually a null node) defines a
bounding region that encompasses the area occupied by all the children
underneath. These parents can be further grouped together and so on. The top-
level parent occupies the entire extent of the world. This structure helps
tremendously in fast culling. If you reach a node that is not in the scene, then
there is no need to traverse to the objects beneath this node. All the objects below
this node will be culled. For example, in Fig.10.5, ,if we reach nodeA and find
that its extents are not within view, then Spherel and Sphere2 are automatically
culled out as well. The test to check for culling is similar to what we discussed
above.
Many times, culling is not just a simple on-off switch. You can define
different levels of detail for the model. As the model gets closer to the viewpoint,
the more detailed version of the model is used.

NodeA
Boundary: (-5. 5) in X and Z

I Spherel
center:(-0 5,O 5 ) radius = I
II Sphere2
center:(-4,-5)radius = I
1
Fig.10.5: Scene Tree

Bells and Whistles: Adding Fog

In many games of today, fog and mist are added to simulate a more moody scene.
OpenGL provides a primitive capability for rendering atmospheric effects
such as fog, mist, and haze. It is useful to simulate the impact of atmospheric
effects on visibility to increase realism. It also allows the designer to cull out
polygons near the fog limit since they cannot be seen anyway.
OpenGL implements fogging by blending the fog color with the incoming
fragments using a fog blending factor,f,

The function

glFogilGL-FOG-MODE, fogfilter);

establishes the fog filter mode. It can be of type GL-EXP, GL-EXP2 or

GL-LINEAR. GL-LINEAR is the best for fog-rendering mode; objects fade in
and out of the fog much better.
The command

glFogfv(GL-FOG-COLOR, fogcolor);

sets the color of the fog.

 The command

will establish how close to the screen the fog should start. You can change the
number to whatever you want depending on where you want the fog to start. The
command

This tells the OpenGL program how far into the screen the fog should go.
Objects farther than this distance from the camera cannot be seen and hence can
be culled out. The code below defines fog for our game using OpenGL.

GLfloat fogColor[41- {0.5f1 0.5f, 0.5f, 1.Of); 11 Fog Color

glFogiIGL-FOG-MODE, GL-LINEAR); I1 Fog Mode
glFogfvlGL FOG-COLOR, fogcolor); 11 Set Fog Color
glFogflGL h~DENSITY, 0.35fl; 11 How Dense Will The Fog Be
g l ~ o g f l ~1.MI; ~ r ~ ~ ~ I1 ~ Fog
~ Start
~ ~Depth~ ,
glFogflGL FOG END, 25.M); I1 Fog End Depth
gl~nablel~-~k~;

Run your program with fog enabled and watch what it does to enhance the
mood of your game. An image from our game is shown in Color Plate 6.
There are many more bells and whistles you can add. You can have rooms
and dungeons set up for the viewer to explore. Andy himself may not be
stationary-he may be animating about the scene, trying to avoid you. Or, he
may be firing bullets at you! You have to hide behind an object, locate him and
fire back. If you hit him first, you win! Otherwise ...

Summary

In this chapter we have explored how to use camera animation within a 3D

world. We animated the camera in the snowman animation to achieve a smooth
camera pan to zoom into snowy. We have also seen how to use camera motion
to simulate a viewer moving through a 3D world. We captured keyboard input to
define the motion of the camera to develop a cool 3D game.
There is no end to the creativity you can put into your graphics. Explore some
ideas yourself and see how you can implement them.
Chapter 11
Lights, Camera,
Action!

In the last few chapters, we looked into the principles of animation and how to
bring our models to life. In this chapter, we shall put together all that we have
learned to develop our own animated movie using Maya PLE.
We shall go through the development of the movie, just as the pros do-from pre-
production, where the story is developed; to production, where the movie is
developed; to post-production, where the frames are polished up and finally put
onto tape..
By the end of this chapter, you will know all the industry buzz words and secrets!
In this chapter, you will learn
Steps taken to develop a movie-from pre-production to post-production
Animation using Maya
Create a fully rendered movie sequence using Maya

Let us start the production of our movie.

We are ready to make a movie, we have the technical know-how-but wait! What
is the story? As a famous director put it, "Without the story, you have nothing."
No amount of technology can save a bad story. The story is paramount to the
success of a project.
Pre-production is the first step in creating an animated project. It's the step
where the story is identified and storyboarded. How does one get ideas for a
story? As you read through this section, your brain will be buzzing with story
ideas-good ideas are really not that hard to come by. (Great ideas, of course, are
not that easy, and take a lot of time and perseverance).

The Story: Idea is Everything

The story is where you identify the main idea of your movie. It's where you
identify the conflict, the resolution and the punch line of the production. Most
films that really grab you can be summed up in one line, called the pitch:
w Jaws: "Man afraid of water pursues killer shark": horror movie for adults
w Toy Story: "Toys come to life": humorous movie for kids
w A Fish Story: "A fish is thirsty, but water is not enough": humorous movie
for readers of this book.
Try to create a couple of these pitches by asking yourself "What if...?"
For example: "What if we had super power?" (Incredibles), "What if..toys
came to life?" (Toy Story) or "What if a fish wanted to taste martini?" (our
movie). Write down as many as you can as fast as you can. Scribble a couple of
notes down around them. At this point don't self-censor, write everything down
however dumb it seems. It's part of the creative process called brainstorming.
Ideas come from everywhere. Sitting in an empty room isn't going to inspire
you-read a daily newspaper, get out to the theatre or nip down the pub. Don't
push trying to get ideas; they'll come. Consider keeping some sort of journal so
you can safeguard everything from getting lost.
Once you have identified an idea, you need to expand it into a fill-fledged
story. The story needs to be visual (so no introspective characters) and, like any
good story, has to have a beginning, middle and an end. You have to filly exploit
your idea, so if you've got a great beginning, make something else happen in the
middle and then give it a good resolution. The story and the characters have to
go somewhere.
Boil a story down to its basics and what usually happens is this: A character
gets involved in some sort of situation that gives the character an aim. The story
relates how the character works to achieve this aim. Just before the end comes
the "make-or-break" time and the aim is usually achieved in some climactic
finale.
Audience
Just as we learned in Chapter 4, the audience is paramount. Detailed research
goes into identifying the audience for a film and their likes and dislikes.
The characters in the story go a long way toward helping the audience
identify with the film. The character' struggles and aims should be something
with which the audience can relate. For example, consider the characters in Toy
Story-the toys in the movie are commonplace toys that everyone can relate
having playing with. And which kids hasn't wondered if their toys could come to
life.
The great thing about the pre-production stage is that you can talk through
your ideas with others. Tell your story to your target audience. The response you
get-"Why doesn't this happen?", "Couldn't she be a suspect?", "I'm not sure about
the ending, but if they did escape..."-will help your story evolve. You will find
yourself thinking out loud, adding new parts, cutting back on stuff that gets a bad
reception. It's the tradition of oral storytelling: audience response is critical in
refining your plot.
In the end, it all comes down to two things:
1) Do you love the idea? You'd better, because you'll be making the movie.
2) Will your audience love the idea and the characters in the story?
Make a movie people want to watch.

For this book, let us design a movie called A Fish Story. The audience is you,
our readers. We do love the story, and hope you will love it as well.
The story line is as follows:
A fish is swimming in a fish bowl. Adjacent to the fish bowl is a martini in a
glass. The fish decides it's thirsty, and it wants martini, not water! It jumps over
to the martini glass: only to find the glass is filled with water too!

Storyboards
We saw storyboards in Chapter 4. Let us explore them in more detail in this
chapter. The concept of storyboarding is rather simple. Storyboarding is a way
for you to organize and preplan your project before attempting to actually
develop it. The storyboarding phase helps identify the shots of the story in
greater detail. It identifies the key frames of the shot, the layout of the characters,
and the general ambient setting and mood of the scene. It also identifies the
rough timing of the shots.
The storyboard is divided into shots. Each shot has key frames that are drawn
out to represent the essence of that shot. The task is to move your story, from
frame to frame, filling in the details.
You don't have to be an artist to storyboard. If you have ever looked at
storyboards by Hitchcock or Spielberg, you have to admit that they can't draw.
Storyboarding is especially usefbl for complex visual sequences e.g. elaborate
shots or special effects sequences.A detailed timing chart is also made at this point.
For our fish story, we identify five shots as described below. The story boards for
the shots are also shown. The timing is calculated based on the way the story
boards play out. It takes a lot of experience to predict good timing. Trial and error
is part of this process.
w Shotl: Opening shot: the camera pans in to show the fish swimming in
a bowl on a table. A martini glass sits next to the bowl. Shotl ends with a close-
up of the fish wiggling his eyes. He has an idea! We define Shotl to be 70 frames
long -just a little more than 2 seconds.

wiggle

Fig.ll.1: Camera pan and zoom on to fish

w Shot2: Cut to show the martini glass. This was what captured the fish's
attention. Since Shot2 is just a still of the martini glass, we define it to be about
20 frames only.

Y Cut to show martini !

Fig.ll.2: Cut to the martini glass

w Shot3: Camera is back to the fish. The fish raises his eyebrows to indicate
he sees the glass. He gets ready to leap over to the martini glass. Shot3 is about
50 frames long.

Still seeing something Gets ready to leap

Fig.ll.3: cut back to fish who is getting ready to leap

Ct IAPI'FR 11 ' LIGHT,CAMERA,
ACTION 265

Shot4: Camera pans back. The shot shows the fish leaping and reaching
the martini glass. Shot4 is 60 frames long.

Into glass

Fig.ll.4: Fish leaps

Shots: Camera closes in on the fish in the martini glass. Oh no! It's not
a martini; it's only water! Shot5 lasts about 50 frames.

h!? - - - _ _ - --

Oh no ! Only water

Fig.ll.5: Its only water !

Here are a few tips for producing quick-and-dirty storyboards.
1) Keep the area you have to draw small. This allows you to draw much
faster. The pictures become more like doodles than works of art.

Remember, the point is to get an idea of how things will look on screen.
2) Copy a set of storyboard sheets so you don't have to spend all night
drawing screen boxes.
3) Sketch in pencil so you can make changes easily. Then ink in for
photocopying. Feel free to use any medium you are happy with
professional storyboard artists use everything from magic markers to
charcoal.
4. Scribble down short notes about what's happening in the shot (e.g.,
"Bob enters") what the characters are saying ("Is this it? Is this how...")
or what kind of sound effects you want ("Roll of thunder").
5. Number your shots so that they can be quickly referenced to on the
shot list and during editing.

Look and Feel

Based on the storyboards, a more detailed environment is defined. The mood of
the scene is identified to guide the color choices and general lighting schemes.
 Each character is defined in more detail, especially its general size and
proportions in relationship to the other characters. Once the story boards and the
general look and feel have been identified, it is time to actually create the
production.
In this case, we want the fish bowl and martini glass to be on a table. The
setting is inside a home. It is probably evening, since the martini glass is out-
room lamps will provide the illumination for the scene. A mirror in the
background help to enhance the visual appeal of the room.
(Its amazing - we have pre-defined models for the above scene from Chapter
8. We shall reuse them in this chapter.)

11.2 Production
Production is the phase of the project where the actual movie content is
developed. In the CG world, production means: the modeling of the characters
and sets, laying out the shots (composing the world), animating the models
within the shots, and finally lighting the scene to achieve the desired ambience.
In the end, you render the shots in order to see your animation in it's full glory.
Start up Maya. Let's get ready to rumble.

Modeling
Based on the storyboards and character design, the characters and sets of the shot
are modeled. The models can be input into the computer in different ways. Some
models are first sculpted in clay and then input into the computer using 3D
scanners. Some models are directly modeled in the computer using software like
Maya.
Materials can also be assigned to the models at this point.
We already have most of the required models (sized appropriately) from
Chapter 8.
We have created the fish and fish bowl model which you can import in:
Models/bowl.mb and Models/Jish.rnb. Interested readers should try and model
their own versions of the fish character.

Layout
The layout phase involves putting together the models to compose the 3D world.
We saw how to lay out some of the world in Chapter 8. Let us put the fish and
fishbowl into the scene as well.
1) Start up Maya.
2) Load in the world you created in Chapter 8. If you don't have one, load
in our model from Models/world.mb.
3) Remove the second martini glass that we had created-we need to make
room for the bowl!
4) Import bowl.mb.
5) Set selection mode to be hierarchy and select the bowl group.
 6) Translate the bowl to (-3.7 ,3.1, -0.46) and set its scale to be ( 0.7,
0.7,0.7).
7) Set the selection mode to be Object.
8) If not already assigned, select only the outside bowl and assign it the
glass shader that already exists, using the marking menu.
9) Assign the water inside the bowl to the water shader.
10) 1mportfish.mb. We have already assigned materials to the fish.
11) Select the entire fish group and translate it to (-2, 2.75,O).
12) Save your scene in a file called worldwithfish.

Animation
Now comes the h n part! Now we get to actually make each shot of our movie
come alive.
Most animation systems (Maya included) use the frame as the basic unit of
measurement for animation sequences. Each frame is played back in rapid
succession to provide the illusion of motion. The frame rate (frames per second
or Qs) that is used to play back an animation is based on the medium in which
the animation will be played back (for example, film, TV, video game, etc.) We saw
that film requires 24 Qs to depict smooth motion, whereas a TV requires 30 fps.
When you set a key frame (referred to as just key in Maya), you assign a
value to an object's attribute: for example, it's translation, rotation, scale, color,
etc., at the specified frame. When you set several keys for an object with
different values, Maya will interpolate the in-between frames as it plays back the
scene. The result is the movement or change over time of those objects and
attributes-animation!
Let's setup Maya for animation.
1) Select the Animation menu set so that the animation related menu
appears.
2) Select Window > Settings/Preferences > Preferences from the menu bar
In the Preferences window that pops up, click the Settings category and
set the Time option to NTSC(30fp) so your animation will play at the
rate of 30 frames per second. Click the Save button.

At the bottom of the Maya UI is the animation timeline, called the Time Slider
in Maya lingo. Look over the playback controls, as shown in the figure below:
Time Slier Playback Conuds
h

\
PlaybackStan Time Playback End n m e
/
Animation Preferences
Range s l i e r

The Time Slider displays the frames of the animation along a time line. The
key frames that you set for the selected object are displayed as red lines along
the time line.
 The Range Slider controls the range of frames that play when you click the
play button. It displays the start time and end time of the entire animation, as well
as Playback Start Time and Playback End Time in text boxes. The playback
times may vary depending on which shot you want to view.
The box at the top right of the Time Slider lets you set the current frame
(time) of the animation.
The Playback Controls control animation playback. You may recognize the
conventional buttons for play and rewind (return to the start time). The stop
button appcars only when the animation is playing. To find out which operation
a buttonttext-box represents, hold the mouse pointer over it.
The Animation Preferences button displays a window for setting animation
preference settings such as the playback speed.
The above items will have more relevance as you work through this lesson.
In the End Time box, set the End Time to be 250. This sets the length of the
entire animation to bc 250 frames long.

Shot1
First, we will set up shotl, which is 70 frames long. In this shot, we will animate
the fish. For 60 framcs, Mr. Fish will swim till the end of the bowl, turn around
and swim back up again. In the next 10 frames, Mr. Fish raises his eyebrows to
indicate he is eyeing somcthing interesting.
Click the rewind button to go to the start of the playback range. This
changes the current frame to 0 (some systems use 1 as the start frame).
Make sure the selection mode is set to hierarchy.
Select the Fish.
Choose Animate > Set Key from the menu bar. This sets up the first key
frame for the fish at the position defined. Notice the red marker that
appears in the Time Slider at frame 0, known as a tick. Ticks in the Time
Slider indicate the key frames for the currently selected object.
At frame 25, we want Mr. Fish to reach the end of the bowl. Go to frame
A convenient way to do this is to click at the desired frame in the Timc
Slider, or to enter this value in the Current Time box.
With the fish selected, go to the Channels Editor and set it's
Translatex=-5.
Select Animate>Set Key to define this kcyframe. Make sure you see the
tick at frame 25.
For the next 10 frames, we want the fish to turn around. Go to frame 35.
Set the fish's RotateY = 180. Set the key frame by selecting
Animate>Set Key. (There is a shortcut to setting the key frame-just
hitting the "s" button. However we find that this doesn't always work
very well.)
At frame 60, the fish reaches the other end of the bowl. Go to frame 60.
Set the fish's Translatex = -2 and set the key frame.
10. Play your animation by clicking the play button from the
 Playback Controls.
From the three keys you've set, Maya creates motion between the positions.
By default, the animation plays in a loop from frame 0 to 60. The animation
occurs in the active window. If you have a fast computer, you might notice that
the animation plays too fast. By default, Maya plays the animation as fast as it
can be processed. Because this scene is simple, the animation might play faster
than the default rate (30 frames per second). When you render all the frames of
your animation for final production, the animation will be smooth. You can also
drag the mouse back and forth (scrub) in the Time Slider to see the in between
frames as you drag the mouse. This is the most basic aspect of our motion.
Following the principles of animation, we can now enhance the animation. First,
let's put in some secondary motion to make the movement look more believable.

Secondary Motion
The tail of the fish should swish as it swims around.
1. Rewind the animation.
2. Set the selection mode to Object.
3. Pick only the tail of the fish (you can pick it from the
Windows>Outliner window as well, the object is called Tail and is
under the Fish parent object.)
4. At frame 0, set the Tail to have a RotateY=20. This will swish the tail
to one side. Set the key frame.
5. At frame 10, set the RotateY to -20. This will swish the tail in the other
direction. Set the key frame.
6. Repeat steps 4-5 for frames 20,30,40,50 and 60. This will cause the tail
to swish from side to side when you playback the animation.
Our shot now calls for a camera pan to zoom into Fishes expression - what
we call staging. To do this, we actually need to create a new camera that can be
animated. (Maya does not allow the default cameras to be animated.)

Staging
1. Choose Create>Cameras>Camera from the main menu.
2. A new camera will appear.
3. Open the Attributes Editor. In the Attributes Editor, you will see the
transformation attributes as well as a tab section to define the camera
 settings of this camera.
4. Rename the camera as Animationcamera by clicking inside the box
with the camera name.
5 . Set the transforms of this camera to be Translate =( 6., 8, 14),
Rotation=( -10, 30, 0).
6. Choose the four panel layout (if it is not already chosen).
7. We don't need the side view window, so we will replace it with the view
seen by our Animation camera.
8. In the menu bar of the side-view window, choose Panels>Look through
Selected. Since the AnimationCamera was selected, this window will
now show the world from the viewpoint of the animation camera.
9. Click into this window and hit 5 to view the objects in a shaded mode.

10. Play the animation with the new camera view selected.
This will be the camera we use for our movie. The animation of a camera is
similar to animating objects. We need to define its key frames, and Maya will
perform the interpolation for us.
We want the camera to be stationary for the first 30 frames. For this, we need
to set a key frame at frame 30. From frames 30 to 60, we will let the camera
zoom into Mr. Fish.
11. Go to frame30.
12. Make sure the transforms of the Animation camera are still defined as
Translate =( 6, 8, 14), Rotate=( -10, 30, 0).
13. Make sure that the animation camera is the selected object, and hit
Animate>Set Key to define this key frame.
14. At frame 60, define the key frame for the animation camera to have
Translate= (-0.883, 3, 2). Keep everything else the same.
Now play your animation (with the animation camera window selected). You
will see the camera zoom on the fish.

More Secondary Animation

From frames 60 to 70, we wish to show that the fish has seen something, and is
conveying this to the audience by lifting his eyebrows in delight.
1. Go to frame 60. Select the eyeball of the Right eye (REye),by locating
it in the outliner window.
2. Show the Channels Editor.
 3. Set the Translate for the eyeball to be ( -0.55,0, 0.11) so that the fish
appears to be looking in front.. Set the key frame.
4. The fish has seen something. He now looks at us with raised eyebrows.
5 . At frame 70, move the eye ball to look at us by setting a key frame with
the Translate values =(-0.56, -0.0 15, 0.1 1).

6. Pick Eyebrow2 from the Outliner Window. Go to frame 60. Make sure
the transforms are set as Translate=( 0.063 , 0 , -0.109) and Rotate=(O,
16.866,O). Save this default position as a key frame.
7. At frame 70, raise the eyebrow by setting Ty = 0.05. Leave the other
transform as is and set the key frame.
8. Select the AnimationCamera. Set frame 70 as a key frame with its
current transformations. This will ensure that the camera stays still till
frame 70.
Play your animation. You will see the fish looking up in delight. What has he
seen? You can enhance the eye motion even further by making the eyebrow
move up and down a couple of times and by widening the eye. We leave this as
an exercise for the reader.

Shot2
Shot2 is a cut to the martini glass. It extends for 20 frames. Its main purpose in
the film is to make the viewer aware of the fact that it was the glass that the fish
saw.
1. At frame 71, we will cut to the martini glass. To do this, set the
transforms for the camera as Translations=(1.06, 2.436, 5.382) and
Rotations=(O, -30, 0) and save the key frame.
2. Go to frame 90 and save the key frame again to keep the camera
stationary between frames 71 and 90.
Shot3
In shot3, we cut back to the fish. For continuity reasons, the camera position is
the same that we ended at in shotl. Shot3 shows the fish excited and getting
ready to launch out of the fish bowl. It extends from frame 91 to frame 140.
1. At frame 91, we cut back to the fish. Go to frame 91. Set the animation
camera transforms back to Translation = (-0.883, 3,2) and Rotation =( -10,
 30,O). Save the key frame.
2. Go to frame 140 and set this as a key frame with the same transforms to
keep the camera stationary during the shot.
The fish saw it! His eyebrows are wiggling.
3. Pick Eyebrow2. At frame 91, lower the eyebrow by setting Translate
Y = 0 and set this key frame.
4. At Frame 94, lift the eyebrow by setting TranslateY = 0.05. Save the
key frame.
5. Lower, lift and lower the eyebrows at frames 97, 100, and 104.
Let's get Mr. Fish to squash and stretch his body in anticipation of a leap.
6. Change the selection mode to hierarchy and pick the fish (make sure the
entire hierarchy is chosen).
7. Go to frame 111 and save the fish's current transformation as a key
frame.
8. Set a key frame at frame 130, by setting its Rotate=(O, 180 50), and
Scale = (0.7, 1.2,l). Do not change the other transforms. This will squash
the fish.
9. At frame 140, set a key frame by changing the fish's Scale=(1.2,0.8,1).
This will stretch the fish in anticipation of the leap.
10. Play back the shot to view your animation. Again, you are encouraged
to use your creativity in enhancing the motion to make it more believable.

Shot4
The fish is ready to fly. He wants some of that martini and he wants it now! In
this shot, we shall pan back the camera so we can see the entire flight of the fish
from bowl to glass. The shot lasts from frame 141 to frame 200.
1. At frame 141, we want to pan back to view the entire scene and see the
fish leaping.
2. Set frame 141 as a key frame for the animation camera, with its
transformations set as Translate= (6,6,14) and Rotation=(- 1O,3O,O).
3. Go to frame 200, and save the current settings of the camera as a key
frame as well.
Let's make the fish fly!
4. Pick the fish. At frame 141, set a key frame with the fish set at Translate
= (-4.5,2.75,0), Rotate = (0,180,70) and Scale = (1.2,0.8,1)
5. At frame 170, set the key frame for the fish as Translate = (0.5,8,0),
Rotate = (0,180,O) and Scale = (1,1,1).
6. At frame 200, set the keyframe for the fish as Translate = (3,5,0), Rotate
= (0,180,-60) and Scale = (1.2,0.8,1). Almost in the martini glass!

Shot5
Shot5 is our finale. It shows the fish landing in the glass. But did he really get
what he wanted?
For continuity reasons, we start the shot with the fish just about to land in the
glass. Shot5 lasts from frame 201 to frame 250.
1. At frame 20 1, select the fish, and set its transforms as Translate = (3.1,
4.45, 0.47), Rotate = (0,180,-50) and Scale = (0.6,0.,6,0.6). The fish is
about to land in the glass. The fish is a bit bigger than the glass, so
we have to cheat on its size a bit! Set the key frame.
2. At frame 210, the fish has landed in the glass. Set frame 210 as a key
frame with the fish transforms as Translate = (3.44,3.466,0.476) and
Rotate = (-60,120,-50).
3. Oh no! It's only water!
4. Till frame 250, you may want to swish the fishs tail slowly, so the scene
doesn't look completely still.
You can add in hrther animation detail to display the sadness that the fish is
probably feeling at the end result of his effort. His eyebrows drooping or his eyes
going half closed are great ways to convey a mood of sadness. What other
motion can you think of!

Previewing the animation

To play a preview of the movie in 30 Eps, we will use the Playblast option
provided by Maya. Playblast enables you to preview the current animation.
When you run Playblast on an animtion, the Playblast outputs a series of
individual frames into a movie file format such as .avi (Windows), .iff (IRIX and
Linux), and .qt (Mac OS X). This is an option is used for preview only!
From the menu bar, choose Window>PlayBlast>O

1 Edit Help
T ~ m eRange r Time Slider 5 StarltEnd
StartTirne
EndT~rne-1 I -
Vtew /;;
Show Ornaments r -
V~ewer(i Movieplayer r fcheck

vv
Dtsplay S~ze From Wtndow T

~ayliast I Appb I Close

I
By default, Playblast previews the animation using the active view and the
current time range in the Time Slider. In the Playblast window, set the StartEnd
option to be on, and change the End Time to be 140. This will enable us to
preview shot1 to shot3 of the animation.
The default scale is 0.5, which makes the Playblast image resolution one-
quarter the size of the active view. If the movieplayer option is on, Playblast uses
the desktop video playback application specified in your Windows environment
(QuickTime or Realplayer) to view the Playblasted images. If you don't have a
movieplayer, select the fcheck option. This will use Maya's internal movie
playback format.
Click on Playblast. This will create a movie for you at thc appropriate speed.
Watch your movie go! Change the StartIEnd time appropriately to view shots 4
and 5. (Note: We have split up the shots because of size restrictions on our
playback system. Your system may be able to handle the entire 250 frames.)
You can go back and change your animation if you don't like anything that
you see.

Lights
Lights are used for much more than just illumination of the scene. The lighting
in a shot can frame the characters, punch up the main idealmodels of the shot,
and define the mood of the scene. Wow!..and you thought lights were an
afterthought.
In Chapter 8, we defined certain mood lighting that matches what we wish to
portray here. We shall keep much of that same lighting. We add in a couple more
lights to punch up the fish, and the martini glass.
To punch up the main characters, we define three spotlights.The first
illuminates the fish bowl and the martini glass from thc front. A top view of the
layout is shown in the Fig.1 1.6
Two spotlights illuminate the models from the back. The back lighting is
defined so that the reflections of the models in the mirror are illuminated as well.
All three lights have a soft white color with an intensity of 0.5.
It's a good idea to render a few frames from every shot and make sure they
are rcndering as desired. Usually, every shot is fine tuned for lighting and
material settings to make sure it looks just right!
We cheated on the light coming from the lampshade, so that our shadows

I
Fig.ll.6: The top view
came out exactly where we wanted. Another point we noted while rendering our
scenes: the fish was casting shadows in the glass bowl as well as in the water.
This effect was not very pleasing, so we turned off the ability for the glass and

water materials to be able to receive shadows. You can do this by selecting the
object, going to the Attributes Menu and turning off the radio
button for Receive Shadows under the Render Stats option. Once you are
satisfied with the test renders, we can set the frames off into batch render mode.

Render
Finally!! It is time to render all those images. Video resolution requires images
be rendered with a resolution of 640 by 480. Film requires 1K resolution or
higher.
We shall render our images to be 640 by 480. Maya has a batch render mode that
allows you to render the entire animation sequence.
To set up the rendering job, proceed as follows
1. From the shelf, select the Render Globals Window icon to display the
Render Global Settings window. In the Render Global Settings window,
select the Common tab, and then open the Image File Output section.
2. In the Image File Output section, set the following options:
a. File Name Prefix: Enter the namefish. This name will be the base
of the filenames created by batch rendering. On top of the Render
Globals window, you will see the Path: variable. This defines the
exact folder in which the rendered files will be placed. Make a note
of it.
b. FrameIAnimation Ext: Select name.#.ext. This specifies that the
filenames will have the format prefix.frarneNumbe~fileFormat.
For example, batch rendering the entire 240-frame animation will
create fish.OOO1.iff, fish.0002.iff7and so on through fish0240.iff.
c. Start Frame: Enter 0, the first frame of the animation sequence to
be batch rendered.
 L Ovss
-I
d. End Frame: Enter 250, the last frame to be batch rendered.
e. Frame Padding: Enter 4. This causes the frameNumber part of the
filenames to be four digits prefixed with 0s. For example, the
filenames will fish.OOO1.iff tinstead of fish. 1.iff.
f. Set the renderable Camera to be Animationcamera. This is the
camera which we will be rendering.
3. For the remaining options in the Render Globals, you'll use the
settings you defined from Chapter 8. Maya will render using
resolution ( 6 4 0 ~ 4 8 0and
) ~ anti-aliasing quality (Production
Quality) with ray tracing on.
From the menu bar, choose Render>Batch Render to kick off your render job. It
will take a while-maybe even hours-so you take a tea break before you view the
final rendered animation.

The post production phase is when editorial comes in. Editors pick up the
animated frames and compile them together to make the final movie. The sound
track, complete with narration, music, and sound effects, is added into the clip.
If the animation is to be mixed in with live action footage then the live action
footage is digitized and composited with the animated sequence as well.
The key to all of this is non-linear computer editing.
In traditional video editing folks would copy segments of your tape to a
master tape, laying shots in order linearly onto the tape. If you got to the end of
a sequence and decided that you wanted to change the third shot, you would have
 to edit the entire sequence all over again. Grrr!
Non-linear computer editing allows you manipulate movie clips and audio
(animated or otherwise) in the same way that you would add and change words
in a word-processor: simply Cut, Copy and Paste the shots to stitch your movie
together.
If you want to do this at home, you will need to get a video capture card,
which will allow you to sample and store clips of your video to hard disk. You
will need to digitize your clips frame by frame to composite them with an
animated sequence.
Non-linear editing software usually comes bundled with the card. Get a
powerful, flexible package that is easy to work with. Some worthwhile packages
are Adobe Premiere, Ulead Mediastudio, Judgement, EditDV and Avid. .The
software will allow you to edit the audio tracks adding voice overs and
background music. You can then use it to the tape the final clip onto film or video
as desired.
For example, in our fish animation, we could use a live-action footage of
people drinking at a party. The footage would need to be digitized into a
sequence of frames. These frames could then be used as images to be reflected
in our mirror (achieved by compositing the live action image and the animated
image with an appropriate mask). This would give the effect of a room filled with
live people. Care must be taken that the lighting of each of the two scenes blends
with the other to make the composite look believable.

Camera
VCT
+r 7
-;-"L. 1
Firewire

I Analog + /
Fig.ll.7: Non-linear editing

Finally, using the non-linear software, you can add in sounds effects and the final
movie clip can be produced and transferred directly to VHS.
The steps followed by professional editorial departments are analogous to the
ones described above. The exact mechanics of non-linear editing is beyond the
scope of this book-maybe in Part 2!
The final images are stored in the directory specified by the Path variable we
noted earlier. Go to this directory. You will see the individual images of the
11.4 Finally: Our Movie
animation saved as fish.*.iff. Once Maya is done generating all the images of the
animation, double click on the fish.0000 image. This will bring up the Fcheck
program that Maya uses to display images and will display the 0th frame of our
render. Choose File>Open Animation and open fish.000.iff again. This will
display the rendered animation of your film. Enjoy!
ColorPlate shows some sample images from our movie. This animated world
is provided for you as well under Models/worldwithJish.mb.

Summary
In this chapter, we have developed a movie following the same processes that the
professionals use. We went from pre-production, where we defined out fish
story; to production, where we animated the movie; and finally to post-
production. We animated and finally rendered a stunning movie using Maya.
This chapter ends our journey into 3D Graphics. We hope you have enjoyed
your experience and are motivated to create your own 3D graphics, using the
techniques you have learned in this book.
Appendix A
OpenGL and GLUT

What is OpenGL?
OpenGL seems to be as close to an industry standard as there is for creating 2D
and 3D graphics applications. OpenGL provides a library of graphics functions
for you to use within your programming environment. It provides all the
necessary communication between your software and the graphics hardware on
your system. The OpenGL API is a portable API that can be compiled and run
on many platforms.
OpenGL programs are typically written in C and C++.

What is GLUT?
GLUT (pronounced like the glut in gluttony) is the OpenGL Utility Toolkit, a window
system independent toolkit for writing OpenGL programs. It implements a simple
windowing application programming interface (API) for OpenGL. GLUT makes it
considerably easier to learn about and explore OpenGL programming. GLUT is
designed for constructing small to medium sized OpenGL programs. While GLUT is
well-suited to learning OpenGL and developing simple OpenGL applications, GLUT
is not a full-featured toolkit so large applications requiring sophisticated user interfaces
are better off using native window system toolkits. GLUT is simple, easy, and small.
The GLUT source code distribution is portable to nearly all OpenGL implementations
and platforms. GLUT is not open source. Mark Kilgard maintains the the copyright.

What libraries and header files will I need to

compile and link OpenGUGLUT programs?
To compile and link OpenGL programs, you'll need the OpenGL header files and
libraries.
Windows: If you are running Windows 98/NT/2000 then this library has
already been installed in your system. Otherwise download the Windows
OpenGL library from Microsoft at:
ftp:/@p.microsoft.com/softlib/mslfiles/opengl95.
exe
Mac OS X: If you are running MacOS X then download the OpenGL SDK
from Apple at:
http://developel:apple.com/opengl/
LINUXIBSD: If you are running LINUX then download Mesa - the OpenGL
work-alike at:
http://www.mesa3d.org/
UNIX: The OpenGL libraries are available directly from your UNIX vendor

The OpenGL Utility Toolkit (GLUT) is available for download at:

http://www.opengl.org/resources/libraries/glut. html
Note that you only need to download the glutdlls.zip archive, but you may also
want to download the source code and the HTML version of the API.

To run OpenGL programs that take advantage of OpenGL 1.2 or above: you will
need to download vendor-specific OpenGL Drivers for your particular graphics
card. Refer to:
http://www.opengl.org/documentation/spec.
html/
for more details.

Where do I install these files?

Microsoft Windows 95 and above:
If you're using Visual C++ under Windows 9x, NT or 2K, your compiler comes
with include files for OpenGL and GLU, as well as .lib files to link with.
Install g1ut.h in your compiler's include directory, glut32.lib in your compiler's lib
directory, and glut32.dll in your Windows system directory (c:\windows\system
for Windows 9x, or c:\winnt\system32 for Windows NTl2000). In summary, a
fully installed Windows OpenGL development environment will look like this:

File Location

[compiler]\include\g1

[compiler]\lib

[system]
where [compiler] is your compiler directory (such as c:\Program Files\Microsoft
Visual Studio\VC98) and [system] is your Windows 9x/NT/2000 system
directory (such as c:\winnt\system32 or c:\windows\system).

You'll need to instruct your compiler to link with the OpenGL, GLU, and GLUT
libraries. In Visual C++ 6.0, you can accomplish this with the Project menu's
Settings dialog box. Scroll to the Link tab. In the Objectllibrary modules edit
box, add glut32.lib, glu32.lib, and opengl32.lib to the end of any text that is
present.

For UNIX or UNIX-like operating systems:

If you don't find the header files and libraries that you need to use in standard
locations, you need to point the compiler and linker to their location with the
appropriate -I and -L options. The libraries you link with must be specified at
link time with the -1 option; -1glut -1GLU -1GL -1Xmu -1X11 -1m is typical.

For Mac OS X:
Apple provides excellent documentation on how to access header and library
files and code using their OpenGL SDK at:
http://developer,apple.com/opengl/

For Linux and other systems:

You can find excellent documentation for this at:
http://www.mesa3d.org/

More OpenGL Documentation

If you would like some more information, then I would suggest you spend time
browsing the OpenGL main web site at:
http://www.opengl.org:
You will find all related documentation and supporting libraries on this web-site.
Appendix B
Downloading and
running Sample Code
How do you download and install the sample code?
Download the sample code (in zipped format) from:
http://wwwspringeronline.com/O-38 7-95504-6
Most of our code is portable to UNIX or any platform with OpenGL
installed.
Go to the directory where you want to install the sample code.
Unzip the downloaded zip file into this folder.

You should now see the following folders and files installed:

OpenGLL contains the examples referred to in the book, grouped by the

example number that they are used within. For e.g., under the folder:
OpenGL/Examplel-l/you will find the source code for Examplel-1 in the file
Examplel-l.cpp.
OpenGLL contains the code for the utility functions:
Vector utilities: vectoz h
BMP image reading and writing: bmp.cpp, bmp.h
VRML models reading: vrml.h, vrml.cpp
' Interpolation routines: linearinterpolation.cpp,1inearinerpolation.h
cubicinterpolatiomcpp, cubicinterpo1ation.h
OpenGL/ModelsL contains the VRML and Alias models used in our
examples.
OpenGL/ImagesL contains the images used as textures in our examples.

How do you compile and link the programs?

We detail this process for Microsoft Visual C++ compiler. Any other visual
compiler will be similar. UNIXJLinux users will need to compile and link their
programs from the command line or from a MakeFile as appropriate (See
Appendix A).
1. Startup Microsoft Visual Studio
2. Create a Project
To create an Open GL project, we are going to need to modify one of the
standard projects defined by Microsoft's Visual Studios.

Select File > New from the main menu. Next select the Projects tab and select
the Win32 Console Application project. Don't forget to enter the project name
and the path to the project folder.

3. Specify The Settings

Select the Project>Settings from the Main Menu.
Select the Link tab. This allows you to edit some of the linking options of the
compiler.
Select the General category from the drop down list.
Select the All ConJprations option from the list. This will make our settings
valid for both compiling the project in Debug mode as well as compiling the
project in Build mode.
Finally, add the glut32.libY glu32.lib and opengl32,lib to the Object/library
modules box.
Click the OK button to save the changes.

g n a m t e debug info r Ignore all default libraries

Link incrementally Generate mapme
r Enable pmfiling
Common Qplions:
kemal32.libuser32.lib gdi3Z.libwinspool.lib comdIg32.lib A '
sdvap132,libshe1132 lib ole32.liboleaut32.lib uuidlib
odbc32Sib odbccp32.lib /nologo /subsystem:wnsole
r :

4. Add the necessary files to the Project

The next step is to add the required .cpp (or .c) files to our project so we can test
the installation. Select Project > Add to Project>Files from the Main Menu. For
this example, select our file Examplel-4/Examplel-4.cpp.Visual Studios will
automatically add this to your project and save it in your project's folder. Follow
this step to add more files if needed.

5. Compile and Execute

Finally, press Ctrl-F5 to build and execute the program. E x a m p l e l should look
something like this:

Enjoy!
Appendix c
Maya Personal Learning
Edition (PLE)

What is Maya?
Academy Award winning Maya software is the world's most powerfully
integrated 3D modeling, animation, effects, and rendering solution. Maya also
adds to the quality and realism of 2D graphics. That's why film and video artists,
game developers, visualization professionals, Web and print designers turn to
Maya to take their work to the next level.

What is Maya PLE?

Maya Personal Learning Edition is a special version of Maya software, which
provides free access to Maya for non-commercial use. It gives 3D graphics and
animation students, industry professionals, and those interested in breaking into
the world of computer graphics (CG) an opportunity to explore all aspects of the
award winning Maya software in a non-commercial capacity.

Where do I download the soffware?

You can download the software from the following URL
http://www.alias.com/eng/products-sewices/mya/myaqle/

Follow the instructions on how to install the program onto your desktop.
This page also provides more details on software and hardware requirements for
running the program.
Bibliography
What follows is a bibliography in computer graphics. In addition to being a list
of references from the various chapters, it also contains references to books and
journals where the latest research in the field is being published.

ADAM9 1 Adams, L. Supercharged C++ Graphics, Windcrest, 1991

APOD99 Apodaca,A.A., L. Gritz., Advanced RenderMan: Creating CGI for

Motion Pictures, Morgan Kaufmann, 1999

BADL91 Badler, N.I., B.A. Barksy, D. Zeltzer (ed.), Making Them Move:
Mechanics, Control, andAnimation of Articulated Figures, Morgan Kaufman, 1991.

BART87 Bertels, R., J. Beatty, B. Barsky, An Introduction to Splines for Use

in Computer Graphics and Geometric Modeling, Morgan Kaufmann, 1987.

BOVIOO Bovik, A,, Handbook of Image and Video Processing, Academic Press,
2000

COHEOl Cohen, E., R. F. Riesenfeld, G Elber, Geometric Modeling with

Splines: An Introduction, AK Peters, Ltd., 2001.

CUNN92 Cunnighamn, S., N. Knolle, C. Hill, M. Fong, J. Brown, Computer

Graphics using Object Oriented Programming, John Wiley and Sons, 1992.

DEB001 deBoor C., A Practical Guide to Splines, Springer-Verlag, 200 1

DERA03 Derakhshani, D., Introducing Maya 5: 3Dfor Beginners, Sybex Inc, 2003

DER098 DeRose, T., M. Kass, T. Truong, Subdivision Surfaces in Character

Animation, ACM Proceedings of the 25th annual conference on Computer
graphics and interactive techniques, pp.: 85-94

DUTR03 Dutre, P., P. Bekaert, K. Bala, Advanced Global Illumination, AK

Peters, Ltd., 2003

FARR97 Farrell, J.A, From Pixels to Animation, Academic Press, 1997

FOLE82 Foley, J., A. van Dam, Fundamentals of Interactive Computer
Graphics, Addison-Wesley, 1982

FOLE93 Foley, J.D., A. van Dam, S.K. Feiner, Introduction to Computer

Graphics, Addison-Wesley Professional; 1993

FOLE95 Foley, J.D., A. van Dam, S.K. Feiner, J.F. Hughes, Computer
Graphics: Principles and Practice in C, Addison-Wesley Professional; 1995

GLAE99 Glaeser, GH. Stachel, Open Geometry: OpenGL +Advanced Geometry,

Springer Verlag, 1999

GLASS9 Glassner, A.S., An Introduction to Ray Tracing, Morgan Kaufrnann,

1989.

GLAS04 Glassner, A.S., Interactive Storytelling: Techniques for 21st Century

Fiction, AK Peters, Ltd., 2004

GLAS95 Glass, G , B. L. Schuchert, The STL Primer, Prentice Hall PTR; 1995

GOUR71 Gouraud, H., Continuous Shading of Cuwed Surfaces, IEEE

Transactions on Computers, C- 20(6), June 1971, pp. 623-629.

HILL00 Hill, F.S., Computer Graphics Using OpenGL, Prentice Hall, 2000

LAM003 LaMothe, A., Tricks of the 3 0 Game Programming Gurus-Advanced

3 0 Graphics and Rasterization, Sams; 2003

LAVE03 Lavender, D., Maya Manual, Springer Professional Computing Series,

2003

LASS87 Lasseter, J., Principles of Traditional Animation Applied to 3 0

Computer Animation, SIGGRAPH 87, pp. 35-44

LENG93 Lengyel, E., Mathematics for 3 0 Game Programming and Computer

Graphics, Charles River Media; 2003

LISC03 Lischner, R., STL Pocket Reference, O'Reilly & Associates, 2003

MAGN85 Magnenat-Thalmann N., D. Thalmann, Computer Animation. Theory

and Practice, Springer Verlag, Tokyo, 1985

NEWM81 Newman W.M., R.F. Sproull, Principles of Interactive Computer

Graphics, McGraw-Hill, 1981
PAIS98 Pai, S.G, R. Pai, Learning Computer Graphics, Springer Verlag, 1998

PARE01 Parent, R., Computer Animation: Algorithms and Techniques, Morgan

Kaufmann, 200 1

PORT84 Porter T., and T. Duff, Compositing Digital Images, SIGGRAPH 84,
pp. 253-259.

PROS94 Prosise, J., How Computer Graphics Work, Ziff-Davis Press,

Emeryville, CAY1994.

REEV83 Reeves, W.T., Particle Systems-A Techniquefor Modeling a Class of

Fuzzy Objects, SIGGRAPH 93, pp. 359-376.

ROSTO4 Rost, R.J., J. M. Kessenich, B. Lichtenbelt, M. Olano, Open GI

Shading Language, Addison-Wesley Professional, 2004

SHIRO2 Shirley, P., Fundamentals of Computer Graphics, AK Peters, Ltd., 2002

SHRE03 Shreiner,D., M. Woo, J. Neider, Tom Davis, OpenGL Architecture

Review Board, OpenGL Programming Guide: The Official Guide to Learning
OpenGL, Version 1.4, Addison-Wesley Professional, 2003

THOM95 Thomas, F., 0. Johnston, Disney Animation, the Illusion of Live,

Disney Editions; revised edition, 1995

WARRO1 Warren, J., H. Weimer, Subdivision Methods for Geometric Design:

A Constructive Approach, Morgan Kaufmann, 2001

WATT93 Watt A., 3 0 Computer Graphics, Addison Wesley, 1993.

Index of Terms
Pixel 6-10, 12, 14, 16, 18-19, 23,25 Normalize 30, 155
CRT 4-6 Matrix
Resolution Identity matrix 32, 35,44-46, 103,
Screen Resolution 5 Composition 27, 41-43
Color Resolution 7,8 Double buffering 36, 104
dpi 5 Coordinates
Frame buffer 6-7, 14-15, 3 1, 50-52, Homegenous coordinates 4 1-42,
55-57, 59 102, 171
Cartesian coordinates 16 Raster images
Physical coordinates 12 Files 48-49
World coordinates 11-12, 53, 75, 102, BMP format 48
104-105, 107-108, 111-112, 114, 153, Bitmaps 51-53
156,229 Pixmaps 5 1-54
Clipping area 10-12, 15, 18, 36 Overlay plane 56
Viewport 11-12, 17-18, 51, 105-108, Logical operations
115-116 AND 56-57, 59,61
Callback functions 14, 18 NOT 56, 59, 61-62
Rasterize 23 OR 56-57, 59, 61-62
Mid-point Algorith XOR 56, 58-59, 77,
Anti-aliasing 23-25, 21 1, 274 Image processing
Transformations Compositing 57, 60, 61, 275,
Geometric transformations 27, 32 Red-eye removal 60, 62-63
Translation 27, 33-39,41-42,44, 3D
46-47, 102, 104, 110, 120, 171, Coordinate System 82
265,270 Right handed coordinate system 82
Scaling 27, 33, 37-40,42-46, 54, Object coordinate system 118
85, 102-104, 171, 186-187, 189, Object space 108, 118, 120
232 Coplanar 85, 87, 88
Rotation 27, 33, 39-42,44,46-47, Normal vector 86-87, 91, 147, 149-
102-104, 110, 117, 122, 125, 127, 151, 157-158, 160, 175,243,
171, 196, 220, 230, 238, 239, 243, Polygon
265,270 Edge 88,92
Object transformations 32, 36, 109 Convex area 88
Vector Flat 85, 88
Magnitude 30, 83, 85 Front-facing 89, 91
Unit vector 30, 40, 83, 86 Back-facing 89-91, 131-132, 149
Backface culling 90-9 1,96, 130-31 incident ray 134, 139
Models reflected ray 134, 171, 179
Wireframe 92, 129, angle of incidence 134-5, 139
Solid 92 angle of reflection 134
Model file Reflection
VRML 98-102, 124, 127, 150, Specular reflection 134, 142, 149,
155-57, 160,255 179
Surfaces Ambient reflection 137-8
Implicit surfaces 93 Difhse reflection
Pivot point 44, 103, 120-22, 125, 191, Lambert reflectance 139
196,226,244 Emission 135, 144
Local origin 120-22 Texture mapping
Camera position Texture map 129, 152-4, 156-7, 161
Eye 106 Texels 152-4
Viewpoint 106 Environment mapping 157, 159,
Image plane 160,250
Projection plane 105, 111, 113-4 Ray tracing
Projections Forward 178
Planar geometric projection 111 Backward 178
Projectors 106, 111, 113 Subdivision surfaces 163, 176,
Parallel Radiosity
Orthographic 113, 115, 187, 190 Color bleeding 180
Perspective 111-4, 171, 183, 187-8, Join points
194, 196,198 knots 170-2, 174, 193
Transformations Curves
Model 107 Bezier 167, 169
Camera 109,242 Hermite 166,224-5
Projection 108-9 B-splines 163, 166, 169-171
Viewport 106-8, 115 Continuity
Rendering 51, 79, 129, 131, 138, 151- CO 166
2, 155, 160, 171, 177, 179, 184, 199, C1 166
257,273-4 C2 166
Lighting 86, 92, 129, 131, 138, 141, Spline
148-150, 154-5, 160-1, 175, 208,210, uniform nonrational spline 170
212,233,236,264,272-3, 275 non-uniform non-rational spline 170
Shading 129, 130, 133, 150-2, 154, Knots
160, 177,203-4,237 Multiple knots 171
Hidden surface removal Nurbs 163, 166-7, 169, 171-78
z-buffering 132-133 Persistence of vision 2 15
depth buffering 132, 138, Frames 2 15-256
Surface materials 129, 135, 199,202,212 Frames per second
Shaders Fps 2156,265, 271,
Vertex shaders 160-1 Key frames 2 16-18, 220-24, 226-
Law of reflection 28, 230-1, 234,236-41
INDEX
OF TERMS

In betweens 216-7
Tweening 2 16
Animation
Cel animation 2 16
cut 216
Interpolation
Linear interpolation 2 18, 220- 1,
222,225-8
Cardinal cubic spline 224,
Principles of animation
Squash and stretch 23 1-2, 270,
Staging 233,267
Anticipation 234-5, 270,
Timing 235-7,261-2
Secondary action 236
Brainstorming 260
Pitch 247, 260
Storyboards 68,261-2, 264
Object culling 255
Scene graph 256
Color Plate 1:RGB colors and the resultant colors when these colors are mixed

Color Plate 2: Snowy on the Alps

Color Plate 3: Texture Images used to define a cube mapping

Color Plate 4: Cube mapped objects seem like they are reflecting their environment
Color Plate 5: A Ray traced scene showing reflections and refractions

Color Plate 6: A scene from the game on Mars.

Color Plate 7: Scenes from A Fish Story