Browse Books

Reviewer: Jeffrey B. Putnam

Images are everywhere in digital media. Almost every Web page has at least one image, and often many more. Image processing is also everywhere, but image processing to many people summons up only paint programs and image editors; there is much more to image processing than that, though such programs often use the deeper kind of image processing in the background. This book is an introduction to image processing and analysis in the deeper sense, covering the structure of digital images, what color means in such an image, and the basic algorithms involved in image processing. It is intended for advanced undergraduates and graduate students, as well as computer professionals. ImageJ, a Java application and class library, is used as the basic implementation language, and many of the algorithms are translated into Java and ImageJ. This provides readers (and students) with example code that can easily be extended or changed in order to experiment with different algorithms and ideas. The book covers a wide variety of topics in detail, including: types of digital images and their encodings; an introduction to ImageJ; point operations; filters of various sorts; finding edges, contours, curves and corners; finding regions in images; color spaces; Fourier transforms and the discrete cosine transform (DCT); and Image comparison. The appendices cover: mathematical notation, notes on using Java, the basics of ImageJ, and the source code for two of the longer algorithms (also available on the associated Web site). All topics are covered nicely and concisely, with both pseudocode and Java descriptions of many algorithms, and thorough consideration of the mathematics underlying the code. In places, the required mathematical background may be more than some computer science (CS) undergraduates are comfortable with. Two chapters stand out in particular: the one on color spaces, and the one introducing spectral methods. In particular, the latter may have one of the nicest short introductions to the topic I have encountered. It quickly and clearly explains the basis for the Fourier transform (FT), what it does, and, generally, how it is useful. In large part, this is managed by simply not discussing the Fast Fourier Transform (FFT), which is an odd choice, but a very effective one. However, using spectral methods to analyze the effect of filters is not covered, and while full detail is probably beyond the scope of the work, a quick overview would help emphasize the importance of such methods. The text is accompanied by a number of images and before-and-after pairs for given algorithms, as well as clear and well-designed diagrams where appropriate. Good exercises for each chapter are given. While the concise discussions are usually a plus, in a few places they are a bit too abbreviated: in a classroom setting this is probably not a serious problem, but for self study it might be more difficult. As an example, in the chapter on corner detection, the section on the local structural matrix assumes some mathematical background (including partial derivatives, and the trace and eigenvalues of matrices). These sections are rare, and will probably not raise serious difficulties for readers with a reasonable mathematical background. Online Computing Reviews Service