The statistical theory of shape (Q1923766)

This book provides a mathematically rigorous and comprehensive introduction to shape analysis and shape statistics following the basic concept developed by David G. Kendall more than ten years ago. This approach is tightly related to F. Bookstein's ideas of shape analysis and covers the representation and comparison of shapes by `landmarks' and the `Procrustes analysis' as well. Remember first that the shape of an object (figure, data set, or image) is the total of all information that is invariant under Euclidean motions and isotropic rescalings. The shape of many real-life particles and irregularly formed bodies can naturally be assumed as random in the widest sense. As it is pointed out in the preface and in the introductory chapter, shape analysis is a rather complex problem. There exist quite different approaches to describe the shape variation, where the models and mathematical tools are mostly determined by the aim which the researcher has in mind. It is a matter of fact that the last decade has seen a growing interest in the statistical theory of shape and its development is driven by statisticians and probabilists as well as by specialists from several fields of applications (e.g. paleontology, archeology, medical sciences etc.). The main point in Kendall's theory and therefore a primary theme of this book is the representation of shapes on differential manifolds and the statistical consequences of this idea. After sketching the background and the complexity of statistical shape analysis the author (who is a scholar of D. G. Kendall) presents a number of notions and results from Euclidean and differential geometry (Section 2), and treats some basic models of stochastic geometry and some topics of probability theory on manifolds (Section 4). Section 3 is focused on the theory of shape spaces and Section 5 deals with distributions of random shapes. These are the central points of the statistical theory of shapes. The final section discusses some applications of the general results presented before. This book is nearly self-contained and clearly organized. It is highly recommended to those who are interested in a mathematically sound presentation of the basic concepts of this relevant but still less known branch of statistics.