I had the pleasure of visiting the Biometric Technology Laboratory at the University of Calgary in 2004 and 2006. This Laboratory is distinctive among other biometric laboratories—it’s unique focus is synthetic biometrics. This is the focus of the book “Biometric Inverse Problems” written by my colleagues from this laboratory.

The reasons for the intense increase in interest in biometrics are applications such as homeland security, testing of biometric devices, training personnel of biometric-based systems, and modeling the attacks on biometric systems.

 This book is about the inverse problems of biometrics that correspond to the synthesis of biometric information. We widely use inverse operators and transforms, such as inverse filtering, inverse Fourier and Laplace transforms in image processing. We know that a system is invertible, if inputs can be recovered from the output except for a constant scale factor. Unfortunately, these techniques are not acceptable for the inverse problems in biometrics.  The reason is that biometric data are complex structures of statistical nature. It implies that it is possible to construct fingerprints, iris, signature patterns using the corresponding requirements to topology of this data. This paradigm is a key point of the book. It should be noted that any solution to an inverse problem in any field helps better understand the direct problem and can give more benefits, for example, reconstruction of an object in topography. This book is the first compiled work in this direction.

The first impression of this book is that it is written in a reader-friendly style. For example, the complicated multidisciplinary problems are introduced softly, that is, the reader feels that authors care about his or her understanding of the materials. However, this style requires very precise balance between professional and simplified representation. The authors try to keep this balance by including some recommendations for further reading with detailed comments at the end of each chapter.

I have found it reasonable to estimate the usefulness of this book for various communities of researchers, including designers of biometric-based systems, experts in image analysis and biometrics, users of biometric devices, and instructors and students of classes on biometrics.

Designers of biometric-based systems can find here a number of innovative ideas on implementation of generators of synthetic biometric data. For example, in Chapter 3, the reader can find a detailed description of the generator of synthetic signatures. The next generation of polygraphs is introduced in Chapter 5. Note, that some advanced polygraph techniques are adopted in PASS [1]. Another application, the iris pattern design, is widely used in oculist practice [3,4]  (Chapter 6). The authors demonstrate, in particular, techniques for automatic iris synthesis. 

Experts in pattern recognition and image analysis can find several new applications of their knowledge in biometrics. Analysis-by-synthesis approach in image modeling is the bridge between image processing and synthetic biometrics. For example, Chapter 5 on synthetic face design advocates for using morphed faces to improve performance of facial recognition systems.

Several features of this book can be used by experts in biometrics.  In particular, specialists in analysis (direct problem of biometrics), can extend the area of expertise by applying synthesis (inverse problems). This book is aimed at helping them in this.

For users of biometric devices, this book provides particular knowledge in understanding of biometric data.  I believe that professionals from such areas as forensic, justice, border security control and automated banking systems, should be familiar with the approaches that can be used to deceive the defense. For example, synthetic fingerprints (Chapter 4) can be generated so that an automatic system is fooled with the fingerprints of individuals who do not exist. These specialists should know that  handwriting can be generated automatically without a person in the loop. This and other useful knowledge can help users of biometric technologies to keep a reasonable level of confidence in biometric data.

Finally, the book can be useful for students and instructors of classes on biometrics. From a quick look through this book, one can conclude that the material is well structured and illustrated. In particular, examples are short, clear, and well placed; summaries give the quintessence of each chapter; problems are useful for detail study. I found especially useful the recommendations and comments for further reading provided in each chapter.

There are still several unsolved problems of synthetic biometrics. It refers to the concept of acceptability of synthetic data—biometric data can be recognized as synthetic data and can be used by experts in automated identification systems [2].  Also it seems to me that the authors overestimate the effectiveness of Voronoi diagrams (Chapter 7) in the synthesis of topological representations of biometric data. From the other side, the final chapter (Chapter 8) can be recognized as a contribution in the next generation of biometric systems, DNA-based biometrics.

In summary, I think that this book can be recognized as an important event in the biometric community and related areas, including pattern recognition.

BOOKSBOOKSBOOKS

Biometric Inverse Problems

by

S. N. Yanushkevich, A. Stoica, V. P. Shmerko, and D. V. Popel

CRC Press/Taylor & Francis Group, 2005

Reviewed by:  Patrick S. Wang