ISP Seminars

Previous Seminars

Diffusion tensor Imaging and challenges to evaluate spinal repair
Damien Jacobs
17 September 2012
Diffusion tensor imaging (DTI) is a non-invasive promising technique for longitudinal study in the biomedical field. Physician and pharmacists apply this biomarker to evaluate the efficiency of drugs and to improve the analysis of preclinical experiments such as spinal repair. However, interleaved DTI sequence carried out at ultra high field are impaired by motion and off-resonance effects. In this case, this biomarker can not be validated for biomedical applications. The purpose is to suggest a new strategy for a robust acquisition and to confirm this potential biomarker.
Ressources :

-- Monday, 17 September 2012 at 14:00 (45 min.)

High Level Markov Modeling and Markov Random Tree recognition
Jerome Plumat
11 June 2012
Markov Random Field (MRF) modeling is a powerful framework allowing to formulate and to solve very complex imaging problems. This talk presents a particular case of MRF: the High Level MRF with application to root segmentation. This framework enables to formulate features based matching. The structures to recognize are assimilated to Markov Random Trees. A curves formulation aims to reduce the solution space and implement complex metrics. Results will be presented on data base and isolate images.

-- Monday, 11 June 2012 at 14:00 (45 min.)

Computations using UCL's grid engine
Prasad Sudhakar
29 May 2012
As a part of our research work, most of us do intensive numerical simulations which require a lot of computation time and memory. Performing these simulations on local standalone computers have several drawbacks: 1) the computers may not be resourceful enough (computing power + memory) to handle the load, 2) several instances of a single piece of code have to be executed serially, one after the other because essentially there is just one computing resource, and hence simulations may last longer hours, 3) license issues for expensive tools such as Matlab, 4) frequent manual intervention, etc. In this talk we will see how to overcome these obstacles by using the UCL/CISM's computational grid, Lemaitre.

-- Tuesday, 29 May 2012 at 15:00 (45 min.)

Aggregation of Local Shortest Paths for Multiple Object Tracking with Noisy/Missing Appearance Features.
Amit Kumar KC
30 April 2012
This talk addresses the multi-object tracking problem. It assumes prior detections of the targets, and uses a graph-based approach to connect detections across time. As a main fundamental contribution, we introduce an original iterative aggregation strategy, which validates non-ambiguous matching first, based on local hypothesis testing about the target appearance. Specifically, each iteration considers a node, named key-node, and investigates how to aggregate it either with previous or subsequent nodes, assuming that the appearance of the key-node is the appearance of the target. In practice, the aggregation is investigated by computing shortest paths within the key-node neighborhood, and the shortest aggregation path is validated for subsequent iterations of the algorithm only when it is considered to be sufficiently better than alternative aggregation options. The approach is multi-scale in the sense that the size of the investigated neighborhood is increased proportionally to the number of detections already aggregated into the key-node. Two main advantages arise from the proposed strategy. On the one hand, by making a (different) hypothesis about the target appearance at each iteration, our framework can benefit from appearance features that are sporadically available, or affected by a non-stationary noise, along the sequence of detections. Whilst those kind of features are frequent in many practical real-life scenarios, to the best of our knowledge, our work is the first one to exploit them without making any a priori assumption about the possible appearances of the tracked objects. Second, the multi-scale and iterative nature of the process makes it both computationally efficient and effective, which is demonstrated through extensive experimental validations.

-- Monday, 30 April 2012 at 14:00 (45 min.)

How to cheat with statistics? A practical guide to the skeptical reviewers.
Maxime Taquet
17 April 2012
Null hypothesis statistical testing and p-values are used pervasively in research. In some fields of research, such as medicine and neuroscience, a p-value lower than 0.05 may simply be a good enough result to publish. But what does this p-value really mean? What does it rely on? Could we reach such a result by chance only? In this talk, I will present answers to these questions as well as their consequences in terms of the confidence we have in some published results. Finally, I will introduce some alternatives to null hypothesis statistical testing inspired from the Bayesian data analysis literature.

-- Tuesday, 17 April 2012 at 14:00 (45 min.)

Depth stitching for extended field of view for highly curved IOL charaterization
Alexandre Saint
12 March 2012
Multifocal intraocular lenses are implanted in the eye to correct vision deficiencies. They are a viable alternative to spectacles and contact lenses. Some advanced IOLs are made of diffractive steps with different resolving power to allow a vision from near to far distances. But these IOLs also come with highly curved and discontinuous surfaces. It is an industrial challenge to be able to charaterize such IOLs precisely and in a short time.

-- Monday, 12 March 2012 at 14:00 (45 min.)

Principles of computer graphics animation
Christian Van Brussel
27 February 2012
In computer graphics, a single image is generated by using some mathematical model that will simulate the propagation of the light, either in the virtual scene and at the contact of the virtual objects. The main principles and techniques for the simulation of those lighting models have been presented in previous presentations (cfr. a SPS of November 2008 and the presentation at the "Quinzaine des technologies de Charleroi" in September 2011).

-- Monday, 27 February 2012 at 14:00 (45 min.)

On the use of Gabor Filters in SOS Corrected Ultrasound Image Segmentation
Augustin Cosse
12 December 2011
In many surgery centers, reduced quality images are acquired during surgery in order to track the tissue deformation in real time. Among those intraoperative imaging modalities, ultrasounds are the subject of an increasing amount of research and the detection of organ contours in those images thus constitutes a point of interest. A new approach for correcting and segmenting US images will be presented. This approach is based on a speed of sound (SOS) correction followed by Gabor filtering and SVM-based classification of the corrected image voxels. A brief introduction to the physics of ultrasound shall be provided as well.

-- Monday, 12 December 2011 at 14:00 (45 min.)

Elliptic curve cryptograpy in JavaScript with application for e-voting
Quentin De Neyer
28 November 2011
Current browsers offer fairly limited support for performing cryptographic operations on the client-side of web applications. The support of the TLS/SSL protocols enables secure client-server communications, but these protocols can only be useful in settings where the server is trusted by the client. There are numerous applications, however, in which it is not desirable to ask web application users to trust a server. E-voting is one of them and motivated our developement of a new cryptographic library, entirely implemented in JavaScript. The use of the JavaScript engine was the most convenient choice for computing on the client side of web applications: a JavaScript engine is provided with all major browsers. On the one hand, despite tremendous improvements during the last two years, the performance of JavaScript code remains extremely low compared to optimized compiled code executed on the same computer. On the other hand, compared to other slow platforms like smart-cards, browsers offer an amount of memory that is larger by orders of magnitude. Such constraints motivated our study. We chosed to implement elliptic curve operations, using precomputation to accelerate the point multiplication methods. We also studied and optimized the performance of the different field operations that are involved in elliptic curve arithmetic. Our results are promising and the proposed libraries can be used in a wide range of applications besides e-voting.

-- Monday, 28 November 2011 at 10:45 (45 min.)

Image-based PUFs for Anti-Counterfeiting
Saloomeh Shariati
7 November 2011
Random variations that appear on the profile of the physical objects can be exploited to prevent their cloning. We present Physical Unclonable Function (PUF) designs that exploit random variations which are characterized by images. This distinct category of PUFs, called Image-based PUFs are functional to protect many products such as luxury products, pharmaceuticals, clothing, watches, automotive spare parts etc. We present a general model of Image-based Physical Function System. We highlight Image-hashing as the particular component of the Image-based Extraction and present two practical non-adaptive and adaptive image-hashing methods namely Random Binary Hashing and Gabor Binary Hashing. We explore experimentally the security properties, i.e., robustness and physical unclonability of an example of image-based physical function namely Laser-Written Physical Function. The results demonstrate significant effect of the image hashing method and parameters on the ultimate trade-off between robustness and physical unclonability.

-- Monday, 7 November 2011 at 14:00 (45 min.)