ISP Seminars

Previous Seminars

Advanced signal processing and VLSI implementation for wireless communication and biomedical applications
Prof. Daniel Massicotte, U. Québec, Canada
27 July 2015

We present the research projects realized at the "Laboratoire des signaux et systems intégrés" du Groupe de recherche en électronique industrielle de l'Université du Québec à Trois-Rivières. This research laboratory develops advanced methods in signal processing and microsystems to target growing needs strategic areas such as telecommunications systems, biomedical, measurement and control.

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Monday, 27 July 2015 at 11:00 (45 min.)

Compressive acquisition of linear dynamical systems (with application to video CS)
Amirafshar Moshtaghpour
12 May 2015

This talk will analyze the paper A. C. Sankaranarayanan, P. K. Turaga, R. Chellappa, and R. G. Baraniuk, "Compressive acquisition of linear dynamical systems," SIAM Journal on Imaging Sciences 6, No. 4, pp. 2109-2133, 2013. (pdf). Its abstract reads:

Ressources :

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Tuesday, 12 May 2015 at 10:45 (45 min.)

Learning Deep Neural Network
Romain Hérault, Maître de conférences, INSA, Rouen, France (invited talk)
29 April 2015

In this talk we will addresses the problem of learning Deep Neural Network (DNN) through the use of smart initializations or regularizations. Moreover, we will look at recent applications of DNN to structured output problems (such as image labeling or facial landmark detection). - Introduction to supervised learning, why using regularization ? why looking for sparsity ? - Introduction to perceptron, multilayer perceptron and back-propagation - Deep Neural Network and the vanishing gradient problem - Smart initializations and topologies (stacked autoencoders, convolutional neural networks) - Regularizing (denoising and contractive AE, dropout, multi-obj) - Deep architecture for high dimensional output or structured output problems

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Wednesday, 29 April 2015 at 10:45 (45 min.)

Hyperspectral Unmixing
Muhammad Arsalan
1 April 2015

The sensors of spectral imaging devices have often large sizes of pixel in which numerous materials contribute to the spectrum obtained at a single pixel. In most of applications we need to identify the constituent materials present along with their proportions in such mixed pixels. The process by which the mixed pixel is decomposed into a collection of constituent spectra (endmembers) and a set of corresponding fractions (abundance, indicate the proportion of each endmember present in the pixel). It is not an easy task because of its ill-posed inverse problem nature due to model inaccuracies, noise in the observation, environmental conditions, data size and endmember variability.

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Wednesday, 1 April 2015 at 10:45 (45 min.)

Non-parametric PSF estimation from celestial transit solar images using blind deconvolution
Adriana Gonzalez
18 March 2015

Characterization of instrumental effects in astronomical imaging is important in order to extract accurate physical information from the observations. Optics are never perfect and the non-ideal path through the telescope is usually represented by the convolution of an ideal image with the Point Spread Function (PSF) of the instrument. Additionally, the image acquisition process is contaminated by other sources of noise. The problem of estimating both the PSF and the denoised and deblurred image, called blind deconvolution, is ill-posed. We propose a blind deconvolution scheme that relies on image regularization. Contrarily to most methods presented in the literature, the proposed method does not assume a parametric model of the PSF and can thus be applied to any telescope. Our scheme is based on a wavelet analysis image prior model and weak assumptions on the filter’s response. We use the observations from a celestial body transit where such object can be assumed to be a black disk. Such constraints limit the interchangeability between the filter and the image in the blind deconvolution problem. The proposed method is applied on synthetic and experimental data. The PSF is estimated for SECCHI/EUVI instrument using the 2007 Lunar transit, and for SDO/AIA with the 2012 Venus transit. Results show that the proposed non-parametric blind deconvolution method is able to estimate the core of the PSF with a similar quality than parametric methods proposed in the literature. We also show that, if these parametric estimations are incorporated in the acquisition model, the resulting PSF outperforms both the parametric and non-parametric methods.

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Wednesday, 18 March 2015 at 10:45 (45 min.)

Post-Reconstruction Deconvolution of PET Images by Total Generalized Variation Regularization
Stéphanie Guérit
10 March 2015

Improving the quality of positron emission tomography (PET) images, affected by low resolution and high level of noise, is a challenging task in nuclear medicine and radiotherapy. The aim of this talk is to present a restoration method, achieved after tomographic reconstruction of the images and targeting clinical situations where raw data are often not accessible. This method relies on classical convex optimization tools and on inverse problem theory. The recently developed concept of total generalized variation (TGV) is introduced in the problem formulation to regularize PET image deconvolution. Some properties specific to PET imaging (such as positivity and photometry invariance) are also added to the model to stabilize the restoration. Theoretical results will be illustrated by experiments on both synthetic data and real patient images.

Ressources :

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Tuesday, 10 March 2015 at 10:00 (45 min.)

Diffusion weighted imaging challenges the neuro histology: dream or reality?
Damien Jacobs
6 February 2015

Diffusion weighted imaging is largely used in medical imaging to diagnosis neuropathologies on patient. The signal of water diffusion along neurons is anisotropic. In the past 20 years, this signal is commonly estimated by a tensor and defined as Diffusion Tensor Imaging (DTI). The tensor groups the different cellular microstructures (neurons, myelin, astrocyte, microglia,...) in one compartiment. Recently, multi compartiments models have been proposed to take into account the different cellular microstructure and the heterogeneity. These models have been validated to the healthy subject. In this research, these multi compartiments models are assessed to characterize the Wallerian degeneration process in the spinal cord after the unilateral dorsal roots transection.

Ressources :

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Friday, 6 February 2015 at 10:00 (45 min.)

Normalized cuts for unsigned and signed graphs
Amit Kumar
14 January 2015

In this talk, I will first rapidly recall few terminologies of graph theory and then talk about the normalized cuts in a graph due to Shi and Malik. Then, I will discuss of how we can generalize to handle signed graphs, in which the weights can be negative too.

Ressources :

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Wednesday, 14 January 2015 at 10:00 (45 min.)

Analysis prior with redundant dictionaries for Compressed Sensing
Kevin Degraux
17 December 2014

This article presents novel results concerning the recovery of signals from undersampled data in the common situation where such signals are not sparse in an orthonormal basis or incoherent dictionary, but in a truly redundant dictionary. This work thus bridges a gap in the literature and shows not only that compressed sensing is viable in this context, but also that accurate recovery is possible via an ℓ1-analysis optimization problem. We introduce a condition on the measurement/sensing matrix, which is a natural generalization of the now well-known restricted isometry property, and which guarantees accurate recovery of signals that are nearly sparse in (possibly) highly overcomplete and coherent dictionaries. This condition imposes no incoherence restriction on the dictionary and our results may be the first of this kind. We discuss practical examples and the implications of our results on those applications, and complement our study by demonstrating the potential of ℓ1-analysis for such problems. The talk will mainly follow the structure of the paper and present additional comments and insights, notably from the work of Nam et al. about the cosparse analysis.

Ressources :

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Wednesday, 17 December 2014 at 10:00 (45 min.)

Community detection and Role extraction in networks
Arnaud Browet
29 October 2014

Network clustering, also known as graph partitioning, has focused the attention of many researchers during the last decade due to the increasing number of potential applications of network science. However, the amount of collected data makes their analysis a real challenge and dedicated algorithms must be developed in order to comprehend them. In this talk, I will first introduce recent developments on community detection and present a fast and highly parallelizable algorithm which outperforms existing methods in term of computational time. Yet, a community structure is not always representative of the actual distribution of the nodes in a network. For example, bipartite or cycle graphs often do not contain communities and are in fact more closely represented by what is known as a role structure. In a second part of the talk, I will present a pairwise node similarity measure which allows to extract those role structures and demonstrate its performance on synthetic and real graphs.

Ressources :

Shannon Seminar Room, Place du Levant 3, Maxwell Building, 1st floor -- Wednesday, 29 October 2014 at 10:00 (45 min.)