From Gestalt Theory to
A Probabilistic Approach
By Agnes Desolneux, Lionel Moisan, Jean-Michel Morel
Reviewed by: Tim Patterson
This self-described collection of notes used for course material was published by Springer in their interdisciplinary mathematics series. These notes have been assembled into a 273- page book of largely independent chapters with copious references and notes. Opinions about the book will likely be mixed due to varying coverage of the material. From the book title, one expects a grand scope for the material to be covered. However, the authors have chosen to avoid a “peanut butter” approach of covering each topic evenly. Indeed, with the span of material in this text, in-depth coverage would probably require a massive tome or multiple-volume collection. Instead, the authors have chosen to cover some material in depth and others almost in passing. For example, detection of vanishing points is very well developed and presented, while Shannon and Nyquist are almost an aside.
The general format of this book is significantly different than for most computer vision texts, which usually start with an explanation of the physics or domain behind the pixels. Instead, starting with gestalt theory enables the authors to assume a collection of pixels as their starting point. With this starting point, it is the task of the book to consider various patterns in the image as either random or meaningful. This approach allows a general application of the material but does limit the development, since no underlying statistical model for the pixels can be assumed.
The book begins with a very good development of gestalt theory. The theory could be linguistically elusive, however the ample illustrations make the material easy to read and understand. The first six chapters comprise a well-illustrated description of many of the basics of the theory. The number of figures alone doesn't tell the story because most are multipart figures; although the text connecting the figures often does not support the development of the theory. Chapter 1 is titled “Gestalt Theory and Computer Vision”. “From Gestalt Theory to Computer Vision” may have been a more accurate title for the book because throughout the text, the pattern of pixels is far more important than the value of the pixels.
Most of the images used in this book are different from the norm in imagery texts. It was almost refreshing to read an imagery book in which the “Lena” image does not appear until more than halfway through and then only as a counter example. In the latter portion of the book, the authors turn to more traditional imagery with a section on edge detection - or boundary detection in their parlance. They include the Mumford-Shah model and the Canny edge detector. Multiple examples are shown on different images under varying conditions. This leads to a section discussing objections and questions about the techniques. Edge detection is then coupled back to the gestalt theory to provide a natural segue into boundary detection. In the topic of boundary detection, snakes or active contours are compared with “meaningful boundaries.” Perhaps the major point in the comparison is that meaningful boundaries or fully automatic and parameter-less snakes are not really possible.
At this point, the grand tour resumes with a higher flyover on the topic of clusters. A dozen pages are dedicated to the topic on the way to binocular grouping. On this topic the authors dive deep again. Many readers might know the material better as stereo vision. The geometry is covered well with an introduction of epipolar constraints. The epipolar constraints lead to the seven point algorithm for stereo pair matching. The authors have included a very nice discussion and approaches to the problem. The approaches covered are rich in depth and wide in scope. They discuss common problems and solutions to the problems.
Following the excellent section on stereo vision, the authors come full circle returning to a study on detection and its relationship to the Helmholtz principle. This brief study section is followed by a chapter tying in the full Gestalt program. The table in this chapter provides a good point of reference for the gestalt material. The final chapter in the book is titled “other theories, discussion.” For many readers this actually might be a good chapter to begin reading the book as it places the material in context with other developments in the imagery world.
Click above to go to the publisher’s web page where there is a description of the book, a link to the Table of Contents, and sample pages.
Book Reviews Published in
the IAPR Newsletter
Close Range Photogrammetry: Principles, Methods, and Applications
by Luhmann, Robson, Kyle, and Harley
Classification and Learning Using Genetic Algorithms: Applications in Bioinformatics and Web Intelligence
by Bandyopadhyay and Pal
An Approximation Theory Viewpoint
by Cucker and Zhou
Character Recognition Systems—A Guide for Students and Practitioners
by Cheriet, Kharma, Liu, and Suen
Geometry of Locally Finite Spaces
Machine Learning in Document Analysis and Recognition
by Marinai and Fujisawa (Editors)
Numerical Recipes: The art of scientific computing, 3rd ed.
by Press, Teukolsky, Vetterling and Flannery
Feature Extraction and Image Processing, 2nd ed.
by Nixon and Aguado
Digital Watermarking and Steganography:
Fundamentals and Techniques
Springer Handbook of Speech Processing
by Benesty, Sondhi, and Huang, eds.
Digital Image Processing: An Algorithmic Introduction Using Java
by Burger and Burge
Bézier and Splines in Image Processing and Machine Vision
by Biswas and Lovell
Practical Algorithms for Image Analysis, 2 ed.
by O’Gorman, Sammon and Seul
The Dissimilarity Representation for Pattern Recognition: Foundations and Applications
by Pekalska and Duin
Handbook of Biometrics
by Jain, Flynn, and Ross (Editors)
Advances in Biometrics –
Sensors, Algorithms, and Systems
by Ratha and Govindaraju, (Editors)
Dynamic Vision for Perception and Control of Motion
by Polanski and Kimmel
Introduction to clustering large and high-dimensional data
The Text Mining Handbook
by Feldman and Sanger
Information Theory, Inference,
and Learning Algorithms
“Foundations and Trends in Computer Graphics and Vision”
Curless, Van Gool, and Szeliski., Editors
Applied Combinatorics on Words
by M. Lothaire
Human Identification Based on Gait
by Nixon, Tan and Chellappar
Mathematics of Digital Images
by Stuart Hogan
Advances in Image and Video Segmentation
Graph-Theoretic Techniques for Web Content Mining
by Schenker, Bunke, Last and Kandel
Handbook of Mathematical Models in Computer Vision
by Paragios, Chen, and Faugeras (Editors)
The Geometry of Information Retrieval
by van Rijsbergen
Biometric Inverse Problems
by Yanushkevich, Stoica, Shmerko and Popel
Correlation Pattern Recognition
by Kumar, Mahalanobis, and Juday
Pattern Recognition 3rd Edition
by Theodoridis and Koutroumbas
Dictionary of Computer Vision and
by R.B. Fisher, et. Al
Kernel Methods for Pattern Analysis
by Shawe-Taylor and Cristianini
Machine Vision Books
CVonline: an overview
The Guide to Biometrics by Bolle, et al
Pattern Recognition Books
Jul. ‘04 [pdf]