Cambridge University

Cambridge University

The Signal Processing Laboratory at the Engineering Department, Cambridge University has been in existence for 30 years, since the early days of digital signal processing. There are now 5 staff members, 3 research associates and 30 PhD students. Three key staff members are Dr Kingsbury, who specializes in image processing and wavelet methods, and Dr Godsill and Dr. Fitzgerald, who specialize in Bayesian statistical methods, applied both to images and audio signals. Dr Kingsbury's Image Processing Group has recently been developing new and efficient ways of decomposing images into forms which emphasize features and objects at a variety of different scales and orientations. To achieve this, a new Complex Wavelet Transform has been developed within the Group. The groups of Dr Godsill and Dr. Fitzgerald have been studying methods of making optimal decisions about features in a wide variety of signals. These methods have been based on the evolving field of Bayesian statistics, and in particular on the use of Markov Chain Monte-Carlo (MCMC) methods for sampling complex multi-variate probability distributions and the use of Markov random fields (MRFs) for synthesizing the local interactions between image pixels.

The role of this group within MOUMIR is to combine new developments in two key areas of image processing, in order to produce algorithms with leading-edge performance for the demanding tasks of object segmentation and texture analysis. These areas are multi-resolution wavelet analysis and Bayesian decision making. Traditional Bayesian methods can achieve near-optimal performance, but at very high computational cost. By combining these methods with complex wavelet decomposition of the input images, considerable computational savings may be achieved with little or no performance loss. The use of Bayesian inference is a common theme within three of the Network members including Cambridge and this will allow clear interaction with the rest of the network.

William Fitzgerald graduated with BSc (Hons) Physics, MSc Physics of Solids, and PhD Physics from the University of Birmingham in 1974. Since 1990 he has been a University Lecturer in Engineering, and Director of Studies in Engineering and a Fellow of Christ's College, Cambridge. He currently leads the Statistical Methods Group within the Signal Processing Laboratory in the University of Cambridge. He was Principal Organizer of a 6 month Isaac Newton Institute for Mathematical Sciences Program on 'Nonlinear and Non-stationary Signal Processing', from July to December 1998.

Simon Godsill received the PhD degree Electrical Engineering, from the University of Cambridge. Since 1996 he has been a University Lecturer at Cambridge. Prior to that he was a founding member of the ground-breaking audio processing company CEDAR Audio Ltd. (1988-90), He has worked extensively in the field of Bayesian signal processing. In recent years he has made a particular specialty of Markov Chain Monte Carlo (MCMC) methods for the solution of complex Bayesian signal processing problems. He has recently published a book: S.J. Godsill and P.J.W. Rayner. Digital Audio Restoration. Berlin: Springer, ISBN 3 540 76222 -1, 1998.

Nick Kingsbury received the PhD degree in 1974 in Electrical Engineering, from the University of Cambridge. Since 1986 he has been a University Lecturer at Cambridge. He currently leads the Image Processing Group within the Signal Processing Laboratory in the University of Cambridge. He has been a member of the Technical Program Committee of a number of IEEE and EURASIP international conferences. His current research interests include image compression, error-robust source coding techniques, and image analysis techniques. He has published over 80 papers at international conferences and in leading international journals.

Manuel Davy is currently a Research Associate at the Department of Engineering, University of Cambridge. From 1993 to 2000 he was in Nantes as undergraduate, then as PhD, in the Ecole Centrale de Nantes, France. In Sept 2000, he gained the PhD degree in signal processing. His research interest concern audio signal processing, classification, detection, using mainly Time-Frequency analysis and MCMC methods. Most of his publications concern signal classification.

Cian Shaffrey received the BAI, BA degrees in Electrical and Electronic Engineering, from Trinity College Dublin in 1999. Currently, he is in his second year of study for the PhD degree in the Signal Processing Research Group, Cambridge University. He has been working on Content Based Image Retrieval as the basis of his PhD, and has been a researcher on MOUMIR since April 2000. His supervisor is Dr Nick Kingsbury.

Cédric Févotte was born in Laxou, France, in 1977 and lived in Tunisia, Senegal and Madagascar until 1995. He graduated from the french engineering school École Centrale de Nantes in 2000 and then received the Doctorat d'Automatique et Informatique Appliquée from the École Centrale and University of Nantes in 2003. Since Nov. 2003 he is a Research Associate with the Signal Processing Laboratory of Cambridge University Engineering Department (UK). His current research interests are statistical signal processing and time-frequency signal representations with application to blind source separation.