Direction des Relations Internationales (DRI)

Programme INRIA "Equipes Associées"
(Demande de prolongation)

I. DEFINITION

NOTA : Si la proposition d'Equipe Associée comporte plusieurs partenaires, français et/ou étrangers, vous pouvez :
- soit ajouter une colonne,
- soit dupliquer le tableau ci-dessus autant de fois que nécessaire, en remplaçant "Coordinateur français ou étranger" par "Autre participant français ou étranger".

La proposition en bref

II. BILAN 2008

Rapport scientifique de l'année 2008

Rapport 2007

Description de l'activité scientifique de l'équipe associée et des résultats obtenus : publications, communications, organisation de colloques, formation, soutenances de thèse, valorisation économique, sociale, industrielle, enregistrement de logiciels, dépôt de brevets ... (1 à 2 pages)

Our team has made a significant progress towards the goals stated in the original proposal. This progress has been a result of continuous discussion and collaboration between the team members. During 2008,our teams have made several visits:

1. Senior researchers Ian Jermyn and Xavier Descombes visited Florida State University from April 5-12, 2008.

2. FSU Prof. Anuj Srivastava visited INRIA Sophia-Antipolis from June 10-15, 2008.

3. FSU Prof. Eric Klassen visited INRIA Sophia-Antipolis from July 2-6, 2008. In addition to discussing research with ARIANA members, he also gave a seminar on July 2.

4. FSU Asst. Prof. Adrian Barbu visited INRIA Sophia-Antipolis from July 7-11.

5. PhD student Aymen El Ghoul will visit Florida State University from December 3-22, 2008.

6. Post-doc Csaba Benedek will visit Florida State University from December 6-20, 2008.

7. PhD student Maria Kulikova will visit Florida State University from February 23 to March 1, 2009.

The results of these visits and our collaboration are described below.

Scientific Results

Below we describe the scientific work performed under several subheadings.

Stochastic Geometry with a Focus on Shape Analysis

We have continued the project concerning the detection of a collection of objects from a satellite or aerial image, using marked point processes. This year we focused on sampling shapes from a given distribution. A code has been developed in C++ which, from a set of learning shapes, computes the mean shape (in the sense of the Karcher mean). The learning shape set is then projected in the tangent space at the point given by the mean shape. We then compute the covariance matrix. Assuming Gaussianity, we then can sample some shapes according to the corresponding distribution. This first step being achieved, we now can define a marked point process where the marks correspond to an element of the shape space. The first model under study is Bayesian. The interaction with the data is then decomposed as a product on each pixel, the potential value depending on whether the pixel belong to a shape or not. The prior model simply consists in penalizing overlaps between neighboring shapes.

Looking for Shapes in Cluttered Point Clouds

We have developed a method for identifying shape classes in point clouds. These clouds contain sampled contour points, which are corrupted by observation noise and the addition of clutter points generated by the background and the object interior. Taking an analysis-by-synthesis approach, we simulate high-probability configurations of sampled contours using models learnt from training data and use them to evaluate the given test data. To facilitate simulations, we have developed statistical models for the sources of (nuisance) variability in the problem: (i) shape variation within classes; (ii) variability in the sampling of continuous curves; (iii) pose and scale variability; (iv) observation noise; and (v) points introduced by clutter. Samplings of a finite number of points from a closed curve are represented by positive diffeomorphisms of the unit circle. We derive probability models on these functions using their square-root forms and the Fisher-Rao metric. Using a Monte Carlo approach, we simulate configurations from a joint prior on the shape-sample space and compare them to the data using a likelihood function. Average likelihoods of simulated configurations lead to estimates of the posterior probabilities of different classes and, hence, Bayesian classification. Currently, we are developing an application of this approach to the classification of objects in remotely sensed images when only a set of primitives (points, lines, and arcs) associated with the objects is given.

Publications

Conference

1. An Efficient Representation for Computing Geodesics Between n-Dimensional Elastic Shapes, S. Joshi, E. Klassen, A. Srivastava, and I. Jermyn, in Proceedings of IEEE Computer Soc. Conf., CVPR, Minneapolis, Minnesota, USA, June 2007.

2. Riemannian Analysis of Probability Density Functions with Applications in Vision, A. Srivastava, I. Jermyn, and S. Joshi, in Proceedings of IEEE Computer Soc. Conf., CVPR, Minneapolis, Minnesota, USA, June 2007.

3. Removing Shape-Preserving Transformations in Square-Root Elastic (SRE) Framework for Shape Analysis of Curves, S. Joshi, E. Klassen, A. Srivastava, and I. Jermyn, in Proceedings of EMMCVPR, Ezhou, China, August 2007.

4. Tree Classification Using Radiometry, Texture and Shape Based Features, M. S. Kulikova, M. Mani, A. Srivastava, X. Descombes, and J. Zerubia, in Proceedings of EUSIPCO 2007, Poznan, Poland, September 2007.

5. Diagramme de phase d'une énergie de type contours actifs d'ordre supérieur : le cas d'une barre longue, A. El Ghoul, I. Jermyn, and J. Zerubia, in Proceedings of RFIA 2008, Amiens, France, January 2008.

6. Automatic flamingo detection using a multiple birth and death process, S. Descamps, X. Descombes, and J. Zerubia, in Proceedings of IEEE Signal Processing Soc. Conf., ICASSP, Las-Vegas, Nevada, USA, March 30 - April 4, 2008.

7. Phase diagram of a long bar under a higher-order active contour energy: application to hydrographic network extraction from VHR satellite images, A. El Ghoul, I. Jermyn, and J. Zerubia, in Proceedings of ICPR 2008, Tampa, Florida, USA, December 2008.

8. Bayesian Classification of Shapes Hidden in Point Cloud Data, A. Srivastava and I. Jermyn, 13^th IEEE Digital Signal Processing Workshop, Marco Island, FL, January 2009 (accepted for presentation).

Journal

1. Looking for Shapes in Two-Dimensional Cluttered Point Clouds, A. Srivastava and I. Jermyn. IEEE Transactions on Pattern Analysis and Machine Intelligence, 04 Sept 2008. http://doi.ieeecomputersociety.org/10.1109/TPAMI.2008.223

2. Square-Root Elastic Representations for Analyzing Shapes of Curves in Euclidean Spaces, S. Joshi, E. Klassen, A. Srivastava, and I. Jermyn. In preparation, September 2008.

Training

The following students are currently working within the Associated Team:

INRIA : A. El Ghoul (PhD),  A. Gamal Eldin (PhD), G. Gherdovich (PhD), M. S. Kulikova (PhD) and C. Benedek (Post-doc).

FSU : S. Joshi (PhD), W. Liu (PhD), and J. Su (Master).

Rapport financier 2008

Justifiez en quelques lignes l'utilisation des crédits et en particulier une utilisation partielle du budget alloué.

(*) Ajouter ou supprimer des lignes au tableau ci-dessus de façon à faire figurer tous les organismes ayant contribué au financement de l'équipe associée

Bilan des échanges effectués en 2008

1. Chercheurs Seniors

(1) DR / CR / professeur
(2) colloque, thèse, stage, visite....
(3) précisez l'unité (mois, semaine..)

2. Juniors

(1) post-doc / doctorant / stagiaire
(2) colloque, thèse, stage, visite....
(3) précisez l'unité (mois, semaine..)

III. PREVISIONS 2009

Programme de travail

Stochastic Geometry with a Focus on Shape Analysis

Statistical Shape Analysis of Two-Dimensional Surfaces

Detection of Objects in Hyperspectral Imagery

Detection and recognition of objects using data from hundreds of frequency bands is very challenging. Consider the following simple example. An object present in an imaged scene renders a distinct signature in each frequency band of the hyperspectral data. In case of shape-based characterization of objects, the shape associated with the object may be different in different frequencies. One would like to use this ensemble of shapes to jointly characterize the original object. The tools from statistical theory of shapes help us define and analyze statistics such as means, covariances, moments and stochastic processes on shape spaces, and therefore can be used in analyzing hyperspectral ensemble of shapes. Consider the problem of detecting lakes using aerial images collecting in multiple frequencies. It is possible that the boundaries of lakes may be hard to detect and classify in visible images, but are easier to analyze in thermal images, especially when the ground is at a higher temperature compared to the lake water. Thermal imagery is very useful in identifying bodies of water, roads, plantations, etc due to differences in their thermal profiles from the surroundings.

Programme d'échanges avec budget prévisionnel

1. Echanges

Décrivez les échanges prévus dans les deux sens : invitations de chercheurs de votre partenaire et missions INRIA vers votre partenaire ;
Précisez s'il s'agit de chercheurs confirmés ou de juniors (stagiaires, doctorants, post-doctorants) ;
Motivez, si possible, les raisons scientifiques (travail commun, workshop,..) et précisez la durée prévue ;
Indiquez les étudiants impliqués dans la proposition. Donnez une estimation de leur nombre, pour chaque partenaire, et précisez si des thèses en cotutelle sont prévues ;
Résumez ensuite ces informations dans les tableaux 1 et 2 ci-dessous en faisant une estimation budgétaire :

Ariana and FSU will jointy organize a one day workshop on "Shape Analysis" in June 2009 at INRIA Sophia Antipolis Méditerranée.

2. Cofinancement

Cette coopération bénéficie-t-elle déjà d'un soutien financier de la part de l'INRIA, de l'organisme étranger partenaire ou d'un organisme tiers (projet européen, NSF, ...) ?
Indiquez ces éléments et donnez les montants associés.

Oui de FSU via des contrats (Army et NSF) pour un montant de 10 000$ (soit 7 359 € au taux du 15/10/08).

3. Demande budgétaire

Indiquez, dans le tableau ci-dessous, le coût global estimé de la proposition et le budget demandé à la DRI dans le cadre de cette Equipe Associée.
(maximum 20 K€ pour une prolongation en 2e année et 10 K€ pour une 3e année).

Remarques ou observations :

Pour 2008, il y a un dépassement de 114,45€ sur le budget attribué à l'Equipe Associée, qui sera pris en charge sur le budget Ariana s'il s'avère que les dépenses effectives (en Décembre 2008) correspondent au budget prévisionnel.

© INRIA - mise à jour le 15/08/2008