Scalable Precision Computational Imaging : from Astronomy to Medicine

Decanato - Facoltà di scienze informatiche

Data: 6 Novembre 2017 / 15:30 - 16:30

USI Lugano Campus, room SI-008, Informatics building

Speaker: Yves Wiaux
  Heriot-Watt University, UK
Date: Monday, November 6, 2017
Place: USI Lugano Campus, room SI-008, Informatics building (Via G. Buffi 13)
Time: 15:30-16:30

 

Abstract:

Important imaging modalities in science and technology rely on sampling images in transform domains rather than in the pixel domain, raising an inverse problem for image recovery. “How to sample data efficiently” and “how to solve the inverse problem for image recovery” are fundamental questions in this field of data science that one can characterise as “computational imaging”. In a precision science perspective, fast acquisition, high resolution, and high sensitivity represent important challenges. Image recovery must thus be scalable to extreme data volumes and accompanied with tools for uncertainty quantification around the proposed estimate.  While optimisation provides scalable algorithmic structures for point estimation, full Bayesian inference is not really scalable but provides the most complete approach to estimation, including uncertainty quantification. Our vision in this talk is that the most powerful methods will emerge at the intersection of these theories. We will discuss recently proposed ideas for parallel and distributed computation, data dimension embedding, calibration, and uncertainty quantification, with application to scalable precision imaging in astronomy and medicine.

 

Biography:

Professor Yves Wiaux received a Ph.D. degree in Theoretical Physics from the Université catholique de Louvain (UCL) in Belgium in 2002. He was a Senior Researcher at the Signal Processing Laboratories of the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland until 2013, where he founded the Biomedical and Astronomical Signal Processing (BASP) group. In 2013 he moved as an Associate Professor, and Professor since 2016, at the School of Engineering and Physical Sciences of Heriot-Watt University in Edinburgh, where he currently runs the BASP group. He is also an Academic Guest at EPFL and an Honorary Fellow of the University of Edinburgh. Prof. Wiaux's multidisciplinary research expertise lies at the intersection between inverse problems and their application to computational imaging, with particular emphasis on medicine and astronomy. Underpinning theories are compressed sensing, optimisation, Bayesian inference, and machine learning. 

 

Host: Prof. Antonio Carzaniga