The introduction describes this book as three separate books. The first is the overview, composed of the figures and their captions (which are long and detailed); the second is the basic computer vision text; and the third is a more advanced treatment that includes the appendices and footnotes (which can be ignored when reading the basic text).

For the purposes of this book, computer vision includes automatic deduction of structure and properties of a (possibly dynamic) three-dimensional world from one or more two-dimensional images of the world. This definition omits work in high-level analysis and heuristic techniques. In order to simplify the discussion, color is also omitted. The book reads like a tour of the literature, with numerous references going back to Euclid; thus, the title is appropriate.

Nalwa makes a real attempt to make the figures stand alone with long and complete captions. This figures-only approach gives a good general introduction appropriate for the casual reader. Some topics are covered in more detail than others in this overview, partly because they rely on the figures for the more complete discussion. This is especially noticeable in the sections on three-dimensional projections and geometry. This overview gives a quick introduction to computer vision with an indication of why it is difficult and a brief description of some of the important results in the field.

In more detail, the introductory chapter introduces the subset of computer vision covered by the book. This chapter also introduces examples from human vision, including optical illusions and a description of the physiology of the eye and the human visual system.

The second chapter describes the image formation process for the computer. It covers the theory behind projections (converting the three-dimensional world into a two-dimensional image) and introduces distortions relating to using real lenses and cameras. The chapter concludes with a description of digital image formation, which includes sampling, quantization, and convolution. The appendices to the chapter detail the Fourier transform and briefly discuss probabilities. The reduction of aliasing to the jagged appearance of oblique lines is not the usual description and does not give the reader the sense that it is a potentially serious problem.

Chapter 3, the longest, treats edge detection and image segmentation. One or both of these initial steps is crucial to move from the raw image to the description of the scene in more symbolic terms.

The next five chapters describe techniques and results in different areas of active research in determining three-dimensional shape from images. The first, line drawing interpretation, is a minor break from the exclusion of high-level vision.

Chapter 5 introduces the mathematical formulations for extracting the three-dimensional shape of an object from the variation in shading. The next chapter discusses the properties of texture in images and how they are used to determine the shape of the world.

Stereo, the use of two or more views to extract three-dimensional descriptions, is the topic of the seventh chapter. First the author presents the geometric theoretical basis for extracting three-dimensional images, then he presents several systems that perform the task.

In computer vision, motion analysis includes techniques similar to stereo analysis (matching features through a series of images and generating three-dimensional descriptions) and motion-field techniques (sometimes called optical flow). Chapter 8 is restricted to motion-field techniques.

The last chapter on techniques explores techniques for representing three-dimensional shapes. First Nalwa gives a set of desirable attributes for representations (they should be sufficient, unique, generative, local, additive, stable, and convenient). Then he presents a variety of popular methods (surface patches, symmetries, generalized cylinders, geometric, Gaussian image, and aspects) that have some of these properties.

He closes with some concluding remarks and a list of approximately 400 references. The book contains a citation index with references distinguished among figures, text, and footnotes, and giving their location in the reference list. The subject index also distinguishes among these three categories.Nalwa carefully researched the references, covering similarities and earlier work far better than most papers in computer vision.

Readers will come away with a sense of some of the difficulties encountered by computer vision systems and a sense of some of the current solutions. While the book is a good overview, it has too few details to stand alone as a graduate text in computer vision.