“Comment on the validity of the statement that ‘Industrial machine vision and image understanding have nothing in common’” is both the first and the last of this book’s end-of-chapter exercises. That discussion question captures the overall goal of the book. It is intended as a fourth-year or early graduate-level text that introduces the basics of machine vision as it might be used in industrial applications.

Since this work is introductory, the three chapters after the introduction are devoted to the basics of image sensors, image representations, and image processing. Vernon does not intend to cover these in the detail found in books on image processing, but this introduction provides the foundation for the later discussions on other analysis techniques.

Chapter 5 discusses the problem of segmentation, or a grouping process where components of a group are similar in some feature. Of the two major kinds of segmentation processes (region-based and edge-based), most of the chapter is on the edge-based techniques, with only a few passing references to region-based segmentation approaches. In the context of industrial applications, this approach is reasonable. The basics of edge detection are introduced, and several different techniques are discussed. Their results are shown on an example image (a tray of wires), which is used later in the robot programming example.

Chapter 6, “Image Analysis,” introduces basic matching techniques for recognition of simple two-dimensional objects. The chapter discusses template matching and statistical pattern recognition for finding patterns in an image. The Hough transform is described for finding curves of a given shape (such as straight lines, circles, and parametric curves).

The seventh chapter introduces shape descriptions for two-dimensional objects, based primarily on the taxonomy of shape descriptions introduced by Pavlidis [1]. This taxonomy classifies shape descriptions according to the use of internal or external properties (the region itself, or the boundary of the region) and according to the use of scalar transforms or space domain techniques. Scalar transforms are computed directly for the feature (length, size, intensity, moment, area, and so on). Space domain techniques include the spatial organization of the boundary and internal descriptors such as the medial axis transform and smoothed local symmetries.

Robot programming is presented in chapter 8. This chapter also discusses three-dimensional coordinates and homogeneous transformations. The complete example includes a robot program to pick up and crimp wires, vision algorithms to recognize the ends of wires, and techniques for relating image positions to the real world (and to the camera or robot coordinate systems). This chapter also describes a means of extracting three-dimensional descriptions using structured light.

The final chapter introduces techniques of image understanding, especially David Marr’s bottom-up analysis concepts (from the image to descriptions, as opposed to top-down approaches, which use strong models of the scene to guide the analysis).

This book is not a detailed reference on algorithms for image processing or image analysis. It is intended as a text and provides pointers into the literature (both journal papers and earlier books) for more information. The references are not always obvious--the introductory chapter has a list of general references, but the chapter makes no direct reference to any of them. Each chapter has a set of exercises that include open-ended questions (such as the one given at the beginning of the review) and specific questions relating to the content of the chapter. This feature makes the book more useful for teaching. The book’s limited scope restricts its usage to introductory courses in computer vision.

In addition to the exercises, the book has references at the end of each chapter and a subject index. The use of a single image through much of the book helps relate the different techniques and fits the introductory purpose of the book. Overall, this work is a good introduction that gives a feel for some of the techniques of image analysis and hints at some of the possible applications.