Session list

Abstract
Sparsity and redundancy have been shown to be effective forces in modeling data. In the past decade, numerous applications in various disciplines have harnessed these ideas, leading often to state-of the are results. However, now that our theoretical and numerical understanding of these tools is more stable and mature, it is time to question sparsity-based modeling in general, and seek ways to improve it. In this session we will consider various extensions and variations of sparse and redundant representation modeling, with the hope to map the flaws of the existing approaches, and discuss future steps that our field may take in moving to better handling of data.

Abstract
Compressed sensing has changed the way that we think of the sampling problem. However, in many cases much more is known about the signal class, at least approximately, while physical constraints can limit the types of measurements acquired. This session will explore the roles of: exploiting structured sparsity, sampling strategies and adaptive versus non-adaptive sampling within the broad context of compressed sensing.

Abstract
Optimization is at the heart of many problems in modern signal and image processing, as well as other closely connected areas such as machine learning. In particular, nonsmooth optimization methods have received considerable interest to solve sparse recovery problems (regression, inverse problems, learning, classification, etc.). Sparse recovery algorithms have an extensive literature, all of which can in essence be categorized into three main classes: convex (geometric), greedy (combinatorial), and probabilistic algorithms. The focus of this minisymposium is to give an overview of recent algorithmic advances in nonsmooth optimization methods in signal processing. Session will gather talks by leading experts in the field.

Abstract
This session will gather three renowned speakers in fields of signal processing away from sparsity: machine learning, signal processing on graphs, and computational photography.

Abstract
Many fundamental scientific questions in astrophysics today require high-sensitivity surveys in the radio frequency bands covering large cosmic volumes in sky-coverage and redshift. Due to the physics of cosmic radio emission, this translates to requirements of imaging dynamic range of 1:Million or more using interferometric imaging. The world-wide radio astronomy community is therefore actively engaged in research for the combined challenge of high precision instrumental calibration, high dynamic range wide-band imaging and the resulting High Performance Computing issues for processing data volumes already in the 1–100 Tera Bytes regime.
In this session we will discuss the state-of-the-art algorithms for the various aspects of wide-band direction-dependent calibration, high dynamic range imaging and related high performance computing. The presentations will include work on wide-band imaging in the presence of direction-dependent effects, direction-dependent calibration including atmospheric/ionospheric effects and imaging challenges with aperture- array telescopes. Where possible, an attempt will also be made to highlight the significant differences as well as commonalities between radio interferometric imaging and medical imaging.

Abstract
Optical interferometry is a technique required to overcome the angular resolution limit imposed to monolithic optical telescopes by the laws of diffraction. By combining light from an array of several small telescopes, the imaging power commensurates with the extent of the array rather than with the size of a single telescope. Two major advances have gradually put the focus of optical interferometry on model-independent imaging: the development of instruments that can combine the light from three telescopes or more simultaneously and the development of modern interferometers with long baselines, such as the VLTI (Very Large Telescope Interferometer) or the CHARA (Center for High Angular Resolution Astronomy) Array. These progresses have opened the field to science cases requiring milliarcsecond angular resolution, such as the imaging of main sequence stars, spotted supergiants, fast rotators, star-forming regions and interacting binaries.
Because interferometric measurements only deliver partial information about the brightness distribution of observed sources, image reconstruction constitutes a challenging problem in terms of signal processing. In fact, the quality of interferometric images, once limited by the capabilities of interferometric instruments, is now critically dependent on the performance of image reconstruction algorithms. This issue is especially crucial as imaging fidelity is required for scientific interpretations that, in this field, routinely challenge established theoretical predictions.
In this session, the speakers will present state-of-the art image reconstruction techniques that are currently being developed for application to optical interferometry. The session will cover the main areas of research: reconstructions from multi-wavelength data, the use of sparse representations and Compressed Sensing for regularization, and advances in using Graphic Processor Units to immensely speed up computations.

Abstract
Bayesian techniques have spread out over all astrophysical domains in the last twenty years, especially in cosmology. The first speaker, Mike Hobson, is a pioneer in the use of Bayesian methodologies in astronomy. He will review Bayesian state of the art methods and present some of his last results. Sparsity has recently emerged in astronomy, and first results are extremely encouraging. Jean-Luc Starck will show how sparsity can be very useful for analyzing Cosmic Microwave Background data, as provided by satellite missions such as WMAP and PLANCK. Filipe Abdalla will show how Blind Source Separation methods can help us to derive important scientific results our of multichannel radio-interferometric data.

Abstract
This session will highlight modern sampling approaches for fast high-resolution MR imaging. Single and multiple coil imaging will be considered, in the context of structural, dynamic, or flow applications.

Abstract
Diffusion MRI (dMRI) is the unique Magnetic Resonance Imaging modality able to quantify in vivo and non invasively the average random thermal movement (diffusion) of water molecules in biological tissues such as brain white matter. Using the water diffusion as a probe, dMRI makes it possible to reconstruct white matter fiber pathways and segment major fiber bundles that reflect the structures in the brain which are not visible to other non-invasive imaging modalities. This modern imaging modality, of great interest to neuroscientists and clinicians, has opened a number of challenging problems.
A keynote talk by Denis Le Bihan, credited with developing, refining, and introducing into research and clinical practice the concept of diffusion MRI, will open this session. Denis will introduce this method, help us better understand water diffusion MRI and shed light on what we are looking at when measuring water molecule motion.
In the following 4 technical talks, given by internationally well known experts in computational diffusion MRI, emphasize will be on the presentation of extremely innovative methods recently developed to process and analyse diffusion weighted images with applications to questions of utmost scientific and clinical importance in brain imaging. The 4 talks will have a focus on these tasks, well known to be very challenging due to the complex underlying properties of the diffusion signal. These talks will include recent developments in complexe diffusion modelling from the processing of High Angular Resolution Diffusion Imaging (HARDI) data. HARDI overcomes the limits of diffusion tensor imaging (DTI) and is able to recover complex fiber crossing configurations. These new methods open the possibility of inferring and recovering a more detailed geometric description of the anatomical connectivity between brain areas. Many recent HARDI reconstruction techniques have been introduced to reconstruct the ensemble average propagator (EAP) that captures the diffusion process of water molecules. In the 4 technical talks of this session, HARDI techniques that currently exist on single-shell and beyond will be presented and discussed with a particular focus on the use of Riemann Finsler geometry and dictionary learning based techniques that reconstruct the EAP and its angular part, the orientation distribution function (ODF). From the ODF, it will be shown how one can perform fiber tracking and reconstruct brain connectivity in healthy subjects and patients with brain tumors. In particular, a neurosurgical application will be illustrated. Finally, starting from the fundamentals and based on the fact that the MR signal can be sensitized to self-diffusion of water molecules whose motional history is influenced by the local micro-structure, the last talk will present the computational challenges and essential features of diffusion MR that makes it a powerful probe to characterize tissue micro-structure.

Abstract
This Biomedical signal processing highlight session will feature three of the most outstanding speakers in their fields. The first talk will discuss how mixing acoustics and optics can be used for biomedical applications in the context of multiwave imaging. The second talk will introduce optical coherence tomography. The third talk will be focused on biomedical applications of time-reversal techniques.