Estimation of Visual Motion in Image Sequences

Preface This thesis has been prepared at the IMSOR group of the Institute of Mathematical Modelling (IMM), the Technical University of Denmark, in partial fulfillment of the requirements for the degree of Ph.D. in engineering. The thesis discuss different aspects of the problem of estimating the visual motion given a sequence of images. The treatment of the subject is not exhaustive, nor detailed, but is intended to provide an insight in motion detection by selected theory and examples. A basic knowledge in the areas of image processing and statistics is required by the reader. Acknowledgements During the pursuit of this work I have drawn on a lot of resources of a technical, social, and, economical nature. I would like to express my gratitude to those who have provided these resources. First of all I would like to express my gratitude to Professor Knut Conradsen, and Dr. Bjarne Kjaer Ersbøll for their guidance and encouragement during the course of this work, and for providing the excellent research facilities of our group at IMSOR, now IMM. Furthermore, I wish to thank my colleagues of the image analysis group at IMM for providing a pleasant and inspiring research as well as social environment. The systematic effort of the group to develop and document software within the HIPS-package has proven invaluable for this work. Especially, I would like to thank Allan Aasbjerg Nielsen for his interest, help, and moral support during all phases of my study. I also thank Anders Rosholm, who implemented some to me very useful algorithms on a massively parallel computer under a CAP project. His interest in the problems concerning the algorithm exceeded anything that might have been expected by reasonable standards. During my stay at the Computer Vision Laboratory at the University of Linköping I received very kind treatment. The stay proved very profitable, and has certainly been important for the contents of this thesis. For that I thank Professor Gösta Granlund and his co-workers. I also thank the Nordic Research Network in Computer Vision for financially supporting me during my stay. v vi Image data was provided by the Danish Meteorological Institute. I wish to than Senior Scientist Hans H. Valeur for his interest in my work. I would also like to use this opportunity to thank the friends I have gotten t know at Studenterforeningen in Copenhagen. The course of this study has co incided with a …