Ray Tracing for Undergraduates

1. Introduction

The computer graphics research community has recently renewed its interest in ray tracing, an
image synthesis algorithm that simulates the interaction of light with an environment to generate
highly realistic images (Figure 1). Recent hardware trends and algorithmic developments make
the technique competitive with raster-based algorithms, and some suggest that ray tracing will
begin to dominate interactive rendering in coming years.

Figure 1: Image synthesis using ray tracing. The ray tracing algorithm
supports complex visual effects that are not easily implemented with raster-based
techniques, including depth-of-field, glossy and specular reflections, refraction,
soft shadows, and diffuse interreflection.

At Grove City College, we have mapped the contents of common graduate-level courses in ray
tracing to an undergraduate audience. Students design and implement a full-featured ray tracing
system in a semester-long course that focuses on:

• the essential physics and mathematics,

• software architecture and the impact of design decisions,
• writing efficient object-oriented code, and
• basic algorithm analysis.

The course also affords an opportunity to introduce students to the relevant computer science
literature, both seminal works and recent innovations, throughout the semester.

In this paper, we provide a brief overview of the visibility problem and two competing
algorithms that are commonly used to solve the problem, we detail the course topics and
methodology we have used, and we describe our experience in a pilot course with a small group
of undergraduate students.

2. Background
Despite the current trends in computer graphics research, a thorough study of ray tracing remains
absent from most undergraduate computer science programs. Typical computer graphics courses
focus on raster-based techniques and the corresponding application programming interfaces
(APIs) such as OpenGL or DirectX. The availability of commodity graphics hardware makes
raster-based techniques widely accessible, and these techniques serve as the basis of many
interactive rendering applications, including computer and console gaming. These
characteristics make the study of raster-based techniques and their APIs popular with
undergraduate students.

In contrast, ray tracing is considered by many to be an historic artifact—perhaps relevant as a

research topic in the 1980s, but largely uninteresting today. This perception may stem from the
traditional use of ray tracing as an offline or batch technique: until recently, generating a single
image required either several minutes of computation on desktop systems or extremely
expensive, highly parallel systems found only in large research laboratories. Others believe that
the physical and mathematical underpinnings put the algorithm beyond the reach of
undergraduate students, relegating the study of the algorithm to graduate-level courses.

However, we believe that ray tracing is an ideal vehicle through which to reinforce and apply
fundamental concepts in computer science. To develop an understanding of the algorithm,
students must integrate and extend knowledge in computer graphics, software engineering,
mathematics, physics, and even human perception. These characteristics, together with hardware
trends and recent research results, make ray tracing both relevant and accessible to undergraduate
students.

2.1. Ray tracing versus rasterization

A fundamental problem in computer graphics is the visibility problem: Given a set of three-
dimensional (3D) objects and a viewing specification, determine which lines or surfaces are
visible from that view point. Many algorithms that solve the visibility problem are available in
the literature, but currently the most prevalent are the z-buffer algorithm and ray tracing.

Current graphics hardware is based on the z-buffer algorithm1, which consists of a loop over the
objects in a scene:

for each object N do

for each pixel P through which N might be visible do
compute color cnew and depth znew
if znew < zpixel then
cpixel = cnew
zpixel = znew
The z-buffer algorithm projects an object toward the screen and writes to any pixels covered by
that object both the distance from the object to the view point (the depth or z value) and the color
information, but only if the new z value is less than the current z value associated with a given
pixel (Figure 2).

pixels

view point

z‐buffer

Figure 2: The z-buffer algorithm. Scene geometry is projected toward the

screen, and the z-buffer is used to resolve visible surfaces based on the distance
between an object and the view point. In this case, triangle T2 becomes the object
visible through the highlighted pixels.

In contrast, the basic ray tracing algorithm2 consists of a loop over all of the pixels in an image:

for each pixel P do

find nearest object visible through P

A 3D line query is used to find the nearest object in the parametric space of the ray (Figure 3).

Ray tracing boasts several key advantages over the z-buffer algorithm. First, for preprocessed
models, ray tracing is sub-linear in the number of objects, N; thus, for some sufficiently large
value of N, ray tracing will always be faster than the z-buffer algorithm, which is linear in N.3
Second, the computational kernel of the algorithm performs a 3D line query, and that same
operation can be reused to generate global illumination effects that also depend on a 3D line
query; these effects include shadows, reflection, and refraction.2 Third, ray tracing is highly
parallel and has been demonstrated to have over 91% efficiency on 512 processors.4 This
characteristic, combined with the advent of multicore microprocessors and algorithmic
developments over the past five years, have made ray tracing an attractive alternative for
interactive rendering. Interactive ray tracing algorithms typically trade software complexity for
performance, but impose only modest hardware requirements to obtain interactive frame rates for
complex scenes.5

pixels

view point

ray tracing

Figure 3: The ray tracing algorithm. Rays (3D half-lines) originating at the
view point are traced into the scene, and the object closest to the view point is
found. Here, the 3D line query determines that triangle T2 is visible through the
highlighted pixel.

With the introduction of interactive ray tracing, computer graphics courses must evolve to reflect
the current trends in computer graphics research. Shirley et al.6 have identified several areas in
the design of computer graphics courses on which interactive ray tracing will likely have an
impact:

• Illumination models. Recursive ray tracing integrates visibility and illumination

computations based on a single computational kernel, the 3D line query. Many
common physical effects that are implemented with specialized tricks or hacks in
raster-based pipelines will evolve into their natural illumination computations.
• Perspective transformation and homogeneous coordinates. Ray tracing simulates
perspective naturally, so the mathematic models necessary to simulate foreshortening
will become less important. This process simplifies the traditional transformation
pipeline and eliminates the need for perspective transformation.
• Higher-order surfaces. Ray tracing computes visibility by mathematically
intersecting a line and an object, and it is often straightforward to ray trace
mathematically defined, higher-order surfaces without tessellation. Current emphasis
on object models composed of triangle meshes will begin to shift to models based on
the native representation of higher-order surfaces.
 • Volumetric effects. Ray tracing supports participating media such as aerial fog or
smoke using semitransparent primitives with dedicated illumination models. Here
again, the need for specialized tricks or other hacks will be greatly reduced.

While we agree that the advent of interactive ray tracing will likely have a significant impact on
the structure of traditional computer graphics courses, we believe that the study of the ray tracing
algorithm is an ideal vehicle through which to reinforce and apply fundamental concepts in
undergraduate computer science programs. We have thus mapped the contents of common
graduate-level courses in ray tracing to an undergraduate audience.

3. Course topics and pedagogy

Our undergraduate ray tracing course, currently entitled “Image Synthesis using Ray Tracing”, is
intended to provide the student with an understanding of the details of the algorithm through the
design and implementation of a full-featured ray tracing program. Extensive programming in
C++ facilitates knowledge development in the core areas related to image synthesis using ray
tracing: (1) the physical and mathematical underpinnings of the algorithm, and (2) efficient
system design and implementation. A description from the course catalog is given in Figure 4.

Image Synthesis using Ray Tracing. An introductory course in the design and
implementation of the ray tracing algorithm for computer graphics. Course topics
include: intersection methods for basic and advanced three-dimensional objects;
local and global shading models; acceleration structures for ray tracing; and
introduction to surface, solid, and volume texturing. Prerequisites: Introduction to
Computer Graphics.

Figure 4: Details of the ray tracing course from the course catalog. Our current ray
tracing course is intended for advanced (junior- and senior-level) computer science
students.

3.1. Course outcomes

Upon completion of the course, it is expected that students will be able to:

• Design and implement the basic components of the ray tracing algorithm using
object-oriented programming and a variety of development tools.
• Identify, describe, and analyze the basic components of the ray tracing algorithm,
including ray/surface intersection methods, lighting and shading models, and
advanced rendering effects.
 • Describe and derive the physical and mathematical basis for the ray tracing algorithm,
including ray/surface intersection methods and the geometric model of light transport.
• Locate, analyze, and present articles describing the theoretical and practical research
results concerning the ray tracing algorithm.

By the start of their junior year, our students have completed two semesters of physics (statics,
dynamics, and kinematics), three courses in calculus, a course in discrete mathematics for
computer science, and a semester of mathematical methods for engineers (probability/statistics
and linear algebra). These courses provide the background required to understand the physical
and mathematical underpinnings of the ray tracing algorithm.

Moreover, our students have completed a three-semester programming core, which provides in-
depth experience with object-oriented programming in C++. The students have also completed a
course in data structures and algorithms, and are comfortable with designing, analyzing, and
coding complex data structures and high-performance algorithms. Finally, the students have
completed courses in computer architecture/organization and operating systems, and are familiar
with process/thread management, memory management, and file input/output. These courses
provide the background required for the design and implementation of an efficient ray tracing
system.

3.2. Teaching and assessment methodology

The course consists of traditional classroom lectures and discussion sections that focus on
questions arising from the programming assignments. During the discussion sections,
implementation details and code examples are provided, as is debugging assistance for common
problems. Chapters in the text, research articles, and student presentations also facilitate the
discussion.

Most computer graphics textbooks include at least one chapter on the ray tracing algorithm.
However, these texts are not sufficient to support a semester-long study of the algorithm and its
applications. We currently are aware of two textbooks dedicated entirely to ray tracing that are
suitable for an undergraduate audience:

• Peter Shirley and R. Keith Morely. Realistic Ray Tracing, 2nd edition. AK Peters,
ISBN 1-56881-198-5, 2003.

This text takes a nuts-and-bolts approach to implementing a ray tracing system. Each
chapter tackles a relatively self-contained piece of the overall system, and code
examples are provided at the end of each.
 • Kevin Suffren. Ray Tracing from the Ground Up. AK Peters, ISBN 978-1-56881-
272-4, 2007.

This text offers a more comprehensive view of ray tracing theory and practice.
Topics range from algorithm basics to advanced techniques that would be explored in
the context of a second course on ray tracing. A wide range of code examples are
also included throughout the text.

The Suffern text was not yet available for our initial offering, so we opted for the Shirley text.
However, in future versions of the course, we plan to adopt the Suffern book as the main text and
suggest the Shirley text as a highly recommended (but optional) resource.

Topics from the text are supplemented by readings from the ray tracing research literature. A
partial bibliography of key ray tracing papers that are read and discussed throughout the semester
includes:

• T. Whitted. An improved illumination model for shaded display. Communications of

the ACM, 23(6):343-349, 1980.
• J. Kajiya. The rendering equation. In Proceedings of SIGGRAPH ’86, pp. 143-150,
1986.
• S. Parker, W. Martin, P.-P. Sloan, P. Shirley, B. Smits, and C. Hansen. Interactive
ray tracing. In Symposium on Interactive 3D Graphics, pp. 119-126, 1999.
• I. Wald, P. Slusallek, C. Benthin, and M. Wagner. Interactive rendering with
coherent ray tracing. Computer Graphics Forum, 20(3):153-164, 2001.

These articles detail many of the theoretical and practical aspects of ray tracing, and are
appropriate for advanced undergraduate students with some guidance. Students are expected to
read each article and actively participate in the corresponding discussions. Moreover, students
are periodically required to submit highlighted versions of some articles, as well as a short (one
to two pages) summary of the authors’ contributions. In these cases, guiding questions are
provided to help focus the students’ reading comprehension.

Graded work products for the course consist of weekly programming assignments in which each
student builds his or her own ray tracer from scratch. Students create web pages to display their
results, and the contents of these pages are submitted for grading. Programming assignments are
typically composed of the following elements:

• Required images: Students must duplicate one or more required images that
demonstrate the required functionality. Scene descriptions in a common format are
provided for the required images.
 • Code listing: Students are required to provide a link to the project source code on
their web pages. Students are not graded on the quality of the code or comments,
only on its presence.
• Creative images: Students are typically required to render one or more images that
demonstrate the features of their ray tracer. Again, students are not graded on the
creativity displayed in these images, only on their presence.
• Design decisions: Students are required to describe any design decisions that were
made while implementing the assignment. Students are not graded on their choices,
only on the completeness of the description and justifications.

The required images represent the vast majority of the credit for any particular assignment
(typically 70-80%), whereas the remaining elements are intended to ensure academic integrity
and reasonable thought processes.

The majority of student programming effort occurs early in the semester, as they bootstrap the
necessary infrastructure. For example, the first three programming assignments include:

1. Basic infrastructure. Students construct the basic foundation for their ray tracer.
C++ classes for Point, Vector, Ray, Color, and Image objects are designed
and implemented. Students must provide all of the legal operators for a given class,
including arithmetic operators, dot product, cross product, length, and normalization
routines.

In this assignment, students also implement a simple Sphere class for testing their
infrastructure. The sphere stores its center and radius, and has a single member
function intersects that indicates whether or not a ray hits the sphere. Throw-
away code to generate rays and test the various classes is provided.

2. Ray tracing architecture. Students continue to develop their ray tracing

infrastructure throughout the course of the semester. Students must implement C++
base classes for the Object, Camera, Light, Material, and Background
objects, as well as an intermediate Primitive class (optional).

In this assignment, students also implement derived classes for sphere and plane
primitives, a pinhole camera, point light sources, simple Lambertian surfaces, and a
constant-colored background.

3. Triangle meshes. Students add simple triangle meshes to their ray tracers by
completing the member function definitions for a Mesh class that is provided.

Later assignments typically consist of minor modifications to the existing infrastructure or the
addition of new functionality that requires very few changes to the code developed in earlier
assignments. Examples include the introduction of acceleration structures such as bounding
volume hierarchies or kd-trees, as well as additional material models. Students are encouraged
to follow the basic software architecture presented in class, but they are free to experiment with
alternative designs.

We have developed a self-contained reference implementation that provides efficient solutions to

each of the programming assignments in the course. The reference implementation provides
students with a target for ray tracing performance, and is available to instructors at other
institutions interested in developing their own ray tracing courses (upon request).

All students at Grove City College receive the most current Hewlett-Packard tablet computers
when they enter as freshmen. The current freshmen model, the HP Compaq Tablet PC tc4400
mobile computer, boasts a 2.16Ghz Intel Core2 Duo Processor and 2GB of RAM. These
systems are more than sufficient for generating the required images within a reasonable period of
time, typically just tens of seconds. Computer laboratory facilities with reasonably up-to-date
equipment would also suffice.

4. Experience

We have offered the ray tracing course as a special topics course, which is one way new courses
are piloted at Grove City College. Four senior students enrolled in the course for credit, and
three sophomore students attended the lectures and discussion sections but did not receive
academic credit.

Although special topics courses are not evaluated in the same manner as standard semester-
length courses, feedback directly from the students indicates that the initial offering was well-
received. At the conclusion of the course, one student writes:

Thanks for this study Dr. Gribble, it was very interesting and now I have a cool program
to show off. It's the biggest system I've made since being at Grove City College so I now
feel like all my time coding and in math classes has paid off, I think that's been the
coolest thing about this ray tracing stuff.

Additional feedback from the senior students indicates that:

• Students gained an appreciation of the extensive mathematics background required by

our computer science program. Ray tracing puts many concepts in calculus, linear
algebra, and numerical analysis to practical use in a context that has been described as
“cool” by the students.
• Students gained an appreciation of the natural science courses required by our
computer science program, particularly physics. Often, students reported that the
 physical concepts related to a particular topic (for example, ideal Lambertian
surfaces) made much more sense after implementing the concept in the context of the
ray tracer.
• Students gained an appreciation for the difficulties involved in developing and
debugging a complex software system. The ray tracer was, for many students, the
first experience with a non-trivial code base that had to be designed and written from
scratch.
• Students spent a significant amount of time on the programming assignments
(presumably relative to their other course work), but the results were satisfying. We
did not receive any complaints about the level of effort required by, nor the time spent
on, the programming assignments.

Student performance in the initial offering was exceptionally high. This result is due, at least in
part, to a biased sampling: the four seniors taking the course for credit initiated the effort, and
the sophomores that attended regularly were invited by the instructor. We hope to see similarly
excellent performance in future versions of the course, but a wider range of students will
undoubtedly bring mixed performance.

5. Summary

We believe that a detailed study of the ray tracing algorithm provides an opportunity to reinforce
and apply fundamental concepts in computer science with undergraduate students. We have
mapped the contents of common graduate-level courses in ray tracing to an undergraduate
audience. Students design and implement a full-featured ray tracing system in a semester-long
course that focuses on the essential physics and mathematics, software architecture and the
impact of design decisions, writing efficient object-oriented code, and basic algorithm analysis.
The course also affords an opportunity to introduce students to the relevant computer science
literature throughout the semester.

Ray tracing provides opportunities for more advanced study and undergraduate research. We are
currently planning a second course in ray tracing focused on topics such as sampling theory and
the aliasing problem; advanced global illumination effects such as soft shadows, depth-of-field,
and motion blur; and Monte Carlo methods for simulating the physical transport of light
throughout an environment.

As a small comprehensive college, we hope to demonstrate that ray tracing is accessible to

undergraduate students at a broad range of colleges and universities, both large and small. We
also hope that our experiences are both insightful and useful to other instructors interested in
developing their own ray tracing courses.
References

[1] E. Catmull. A subdivision algorithm for computer display of curved surfaces. PhD dissertation, University of
Utah, 1974.

[2] T. Whitted. An improved illumination model for shaded display. Communications of the ACM, 23(6):343-349,
1980.

[3] J. Cleary, B. Wyvill, G. Birtwistle, and R. Vatti. A Parallel Ray Tracing Computer. In Proceedings of the
Association of Simulat Users Conference, pp. 77-80, 1983.

[4] S. Parker, W. Martin, P.-P. Sloan, P. Shirley, B. Smits, and C. Hansen. Interactive ray tracing. In Symposium
on Interactive 3D Graphics, pp. 119-126, 1999.

[5] I. Wald, P. Slusallek, C. Benthin, and M. Wagner. Interactive rendering with coherent ray tracing. Computer
Graphics Forum, 20(3):153-164, 2001.

[6] P. Shirley, K. Sung, E. Brunvand, A. Davis, S. Parker, and S. Boulos. Rethinking Graphics and Gaming
Courses Because of Fast Ray Tracing. In ACM SIGGRAPH 2007 Educators Program, article number 15, 2007.