Sparsity Averaging for Compressive Imaging
The Faculty of Informatics is pleased to announce a seminar given by Yves Wiaux
DATE: Wednesday, October 3rd, 2012
PLACE: University of Lugano, room A13, Red building (Via G. Buffi 13)
In this talk we will review a novel regularization method recently proposed for sparse image reconstruction from compressive measurements. The approach relies on the conjecture that natural images exhibit strong average sparsity over multiple coherent frames. The associated reconstruction algorithm, based on an analysis prior and a reweighted l1 scheme, is dubbed Sparsity Averaging Reweighted Analysis (SARA). We will probe this prior and the associated algorithm by analyzing the results of extensive numerical simulations for spread spectrum and Gaussian acquisition schemes suggested by the recent theory of compressed sensing with coherent and redundant dictionaries. These results show that average sparsity drastically outperforms state-of-the-art priors that promote sparsity in a single orthonormal basis or redundant frame, or that promote gradient sparsity. We will also illustrate the performance of SARA in the context of Fourier imaging, for particular applications in astronomy and medicine.
Dr Yves Wiaux received the Ph.D. in Theoretical Physics in 2002 from the Catholic University of Louvain (UCL, Belgium). He was a Postdoctoral Researcher at the Signal Processing Laboratories of the Ecole Polytechnique Fédérale de Lausanne (EPFL, Switzerland) from 2003 to 2008. As of today and since 2009, he is a Senior Researcher at the Departments of Radiology of the University of Geneva and of the Lausanne University Hospital, with joint affiliation at the EPFL Signal Processing Laboratories.
Dr Wiaux's expertise and research activities extend from advanced signal processing (inverse problems, sparsity, wavelet techniques, compressed sensing, signal processing on the sphere) to applications in biomedical sciences (data acquisition and image reconstruction in magnetic resonance imaging) and astrophysics (imaging techniques in radio/optical interferometry, cosmological data analysis). He is the author of more than 50 publications in these fields and about the same number of contributed and invited talks. He is also head of the Biomedical and Astrophysical Signal Processing (BASP) group at EPFL, founder of the International BASP Frontiers Workshop, and principal investigator numerous research projects
HOST: Prof. Mauro Pezzč