352.

[Cover] [Contents] [Index]

Page xi

image restoration, robust estimation in the presence of noise (outliers), and the derivation of Markov-based texture measures, are also presented.

Chapter 7 provides several approaches for dealing with multisource data. The methods described include the extension of Bayesian classification theory, evidential reasoning and Markov random fields.

No-one is more aware of a book’s deficiencies and inadequacies than its authors. Even Socrates, after a lifetime of learning, is reported to have been impressed by the extent of his own ignorance. Had more time and space been available, the book would have contained a longer account of the use of wavelets in texture analysis, and of the applications of decision tree classifiers. Publishers, who live in the real world, impose constraints of time and space while authors, naturally, attempt to rewrite their manuscripts every month in order to include the latest developments. We hope that we have reached a happy compromise that should satisfy most readers. We have included references to further work in order to guide the more advanced reader towards the relevant literature.

Most of the research underlying the ideas presented in this book was carried out while Dr Brandt Tso was a postgraduate student, and later postdoctoral fellow, in the School of Geography, The University of Nottingham, under the supervision of Professor Paul M.Mather. The second author provided encouragement, support, contributions to the first three chapters, and numerous rewrites of the draft. We realise that any book written by human authors is necessarily flawed, and we accept responsibility for any errors that may be contained in these pages.

The School of Geography, The University of Nottingham, provided computing facilities, as well as a stimulating and encouraging environment for research. The second author is grateful to his many postgraduate research students from different parts of the world who have, over the past decade or so, educated and trained him in many areas of remote sensing. In particular, he would like to thank Valdir Veronese, Taskin Kavzoglu, Carlos Vieira and Mahesh Pal, who have carried out research projects in areas relevant to the subject matter of this book. The contribution of others, while not directly related to the topic of image classification, has helped by broadening the intellectual debate within his research group, as well as helping in many other ways. The authors also thank Dr M.Koch of Boston University for help and guidance with the Red Sea Hills data set, which is used in a number of examples in this book. The help, good humour and patience of Tony Moore of Taylor and Francis Ltd is greatly appreciated. Richard Willis, of Swales & Willis, helped in numerous ways in the final production stages. Finally, both authors recognise the contributions of their families, and dedicate the book to them.

Brandt Tso     Paul M.Mather
Taipei, 2000     Nottingham, 2000

[Cover] [Contents] [Index]