Page 44 -

1.4 Sample syllabus 23

models is often called image-based modeling or 3D photography. Section 12.6 examines
three more specialized application areas (architecture, faces, and human bodies), which can
use model-based reconstruction to fit parameterized models to the sensed data. Section 12.7
examines the topic of appearance modeling, i.e., techniques for estimating the texture maps,
albedos, or even sometimes complete bi-directional reflectance distribution functions (BRDFs)
that describe the appearance of 3D surfaces.
In Chapter 13, we discuss the large number of image-based rendering techniques that
have been developed in the last two decades, including simpler techniques such as view in-
terpolation (Section 13.1), layered depth images (Section 13.2), and sprites and layers (Sec-
tion 13.2.1), as well as the more general framework of light fields and Lumigraphs (Sec-
tion 13.3) and higher-order fields such as environment mattes (Section 13.4). Applications of
these techniques include navigating 3D collections of photographs using photo tourism and
viewing 3D models as object movies.
In Chapter 13, we also discuss video-based rendering, which is the temporal extension of
image-based rendering. The topics we cover include video-based animation (Section 13.5.1),
periodic video turned into video textures (Section 13.5.2), and 3D video constructed from
multiple video streams (Section 13.5.4). Applications of these techniques include video de-
◦
noising, morphing, and tours based on 360 video.
Chapter 14 describes different approaches to recognition. It begins with techniques for
detecting and recognizing faces (Sections 14.1 and 14.2), then looks at techniques for finding
and recognizing particular objects (instance recognition) in Section 14.3. Next, we cover the
most difficult variant of recognition, namely the recognition of broad categories, such as cars,
motorcycles, horses and other animals (Section 14.4), and the role that scene context plays in
recognition (Section 14.5).
To support the book’s use as a textbook, the appendices and associated Web site contain
more detailed mathematical topics and additional material. Appendix A covers linear algebra
and numerical techniques, including matrix algebra, least squares, and iterative techniques.
Appendix B covers Bayesian estimation theory, including maximum likelihood estimation,
robust statistics, Markov random fields, and uncertainty modeling. Appendix C describes the
supplementary material available to complement this book, including images and data sets,
pointers to software, course slides, and an on-line bibliography.

1.4 Sample syllabus

Teaching all of the material covered in this book in a single quarter or semester course is a
Herculean task and likely one not worth attempting. It is better to simply pick and choose
topics related to the lecturer’s preferred emphasis and tailored to the set of mini-projects
envisioned for the students.
Steve Seitz and I have successfully used a 10-week syllabus similar to the one shown in
Table 1.1 (omitting the parenthesized weeks) as both an undergraduate and a graduate-level
10
course in computer vision. The undergraduate course tends to go lighter on the mathematics
and takes more time reviewing basics, while the graduate-level course 11 dives more deeply
into techniques and assumes the students already have a decent grounding in either vision

10 http://www.cs.washington.edu/education/courses/455/
11 http://www.cs.washington.edu/education/courses/576/