|
|
Xavier Descombes
Researcher, INRIA
 Keywords : Stochastic Geometry, Markov Random Fields, MCMC, Object extraction
 Projects : Mode de Vie (PI), EcoNet (PI), Shapes, Color ODEUR, ARC DADA (PI)
 Demos : see this author's demos
 Contact :
| Mail : | | XavierdotDescombesatinriadotfr | | Phone : | | (33)4-92-38-76-63 | | Fax : | | (33)4-92-38-76-43 | | Postal adress : | | INRIA Sophia Antipolis
2004, route des Lucioles
06902 Sophia Antipolis Cedex
France | | Webpage : | | visit ! |
|
| Abstract :
| My work concerns image analysis using probabilistic models. The first part of my research is devoted to the Markov Random Field approach which includes modeling (the Chien-model,...), estimation (An MCMCML estimator,...) and Optimization (Langevin dynamics,...). The second part concerns the Marked Point Process framework and its application in feature extraction from images (road and hydrographic networks, trees, building,..). |
Last publications in Ariana Research Group :
Building Development Monitoring in Multitemporal Remotely Sensed Image Pairs with Stochastic Birth-Death Dynamics.
C. Benedek and X. Descombes and J. Zerubia. IEEE Trans. Pattern Analysis and Machine Intelligence, 34(1): pages 33-50, January 2012.
Keywords : Building extraction, Change detection, Marked point process, multiple birth-and-death dynamics.
Copyright : IEEE
@ARTICLE{benedekPAMI11,
|
| author |
= |
{Benedek, C. and Descombes, X. and Zerubia, J.}, |
| title |
= |
{Building Development Monitoring in Multitemporal Remotely Sensed Image Pairs with Stochastic Birth-Death Dynamics}, |
| year |
= |
{2012}, |
| month |
= |
{January}, |
| journal |
= |
{IEEE Trans. Pattern Analysis and Machine Intelligence}, |
| volume |
= |
{34}, |
| number |
= |
{1}, |
| pages |
= |
{33-50}, |
| url |
= |
{http://dx.doi.org/10.1109/TPAMI.2011.94}, |
| keyword |
= |
{Building extraction, Change detection, Marked point process, multiple birth-and-death dynamics} |
| } |
Abstract :
In this paper we introduce a new probabilistic method which integrates building extraction with change detection in remotely sensed image pairs. A global optimization process attempts to find the optimal configuration of buildings, considering the observed data, prior knowledge, and interactions between the neighboring building parts. We present methodological contributions in three key issues: (1) We implement a novel object-change modeling approach based on Multitemporal Marked Point Processes, which simultaneously exploits low level change information between the time layers and object level building description to recognize and separate changed and unaltered buildings. (2) To answering the challenges of data heterogeneity in aerial and satellite image repositories, we construct a flexible hierarchical framework which can create various building appearance models from different elementary feature based modules. (3) To simultaneously ensure the convergence, optimality and computation complexity constraints raised by the increased data quantity, we adopt the quick Multiple Birth and Death optimization technique for change detection purposes, and propose a novel non-uniform stochastic object birth process, which generates relevant objects with higher probability based on low-level image features. |
Tree crown detection in high resolution optical images during the early growth stages of eucalyptus plantations in Brazil.
J. Zhou and C. Proisy and X. Descombes and J. Zerubia and G. Le Maire and Y. Nouvellon and P. Couteron. In Asian Conference on Pattern Recognition (ACPR), Beijing, China, November 2011.
Keywords : tree detection, Eucalyptus plantation, Marked point process, multi-date detection.
@INPROCEEDINGS{Zhou11,
|
| author |
= |
{Zhou, J. and Proisy, C. and Descombes, X. and Zerubia, J. and Le Maire, G. and Nouvellon, Y. and Couteron, P.}, |
| title |
= |
{Tree crown detection in high resolution optical images during the early growth stages of eucalyptus plantations in Brazil}, |
| year |
= |
{2011}, |
| month |
= |
{November}, |
| booktitle |
= |
{Asian Conference on Pattern Recognition (ACPR)}, |
| address |
= |
{Beijing, China}, |
| url |
= |
{http://hal.inria.fr/hal-00740973}, |
| keyword |
= |
{tree detection, Eucalyptus plantation, Marked point process, multi-date detection} |
| } |
Abstract :
Individual tree detection methods are more and more present, and improve, in forestry and silviculture domains with the increasing availability of satellite metric imagery. Automatic detection on these very high spatial resolution images aims to determine the tree positions and crown sizes. In this paper, we used a mathematical model based on marked point processes, which showed advantages w.r.t. several individual tree detection algorithms for plantations, to analyze the eucalyptus plantations in Brazil, with 2 optical images acquired by the WorldView-2 satellite. A tentative detection simultaneously with 2 images of different dates (multi-date) was tested for the first time, which estimates individual tree crown variation during these dates. The relevance of detection was discussed considering the detection performance in tree localizations and crown sizes. Then, tree crown growth was deduced from detection results and compared with the expected dynamics of corresponding populations. |
Reconstruction 3D du bâti à partir d'une seule image par naissances et morts multiples.
J.D. Durou and X. Descombes and P. Lukashevish and A. Kraushonak. In Proc. GRETSI Symposium on Signal and Image Processing, Bordeaux, France, September 2011.
@INPROCEEDINGS{DurouGretsi11,
|
| author |
= |
{Durou, J.D. and Descombes, X. and Lukashevish, P. and Kraushonak, A.}, |
| title |
= |
{Reconstruction 3D du bâti à partir d'une seule image par naissances et morts multiples}, |
| year |
= |
{2011}, |
| month |
= |
{September}, |
| booktitle |
= |
{Proc. GRETSI Symposium on Signal and Image Processing}, |
| address |
= |
{Bordeaux, France}, |
| url |
= |
{http://hal.inria.fr/inria-00625527/fr/}, |
| keyword |
= |
{} |
| } |
Résumé :
Dans cet article, nous nous écartons de l’approche classique qui considère la reconstruction 3D comme un problème inverse et la
résout en mettant en correspondance deux images d’une paire stéréoscopique. Au contraire, nous montrons qu’il est plus simple de résoudre le
problème direct. Pour ce faire, nous proposons aléatoirement des configurations de bâtiments pour ne conserver que les plus pertinentes par un
algorithme de type naissances et morts multiples. Nous montrons notamment que cette approche ne nécessite pas un temps de calcul prohibitif,
grâce à la puissance de calcul d’OpenGL qui s’appuie sur la carte graphique. Les premiers résultats obtenus montrent la pertinence de l’approche
adoptée. En particulier, elle permet de résoudre des ambiguïtés pour lesquelles l’inversion du problème serait quasiment impossible. |
Abstract :
In this paper, contrary to the classical approach addressing the 3D reconstruction problem as an inverse problem and solving it by matching two images from a stereoscopic pair, we show that we can solve the direct problem in a simpler way. To do so, we randomly propose configurations of buildings while keeping only the most relevant ones, using a multiple births and deaths algorithm. Notably, we show that this approach does not imply a prohibitory computation time, thanks to the freeware OpenGL which exploits the graphic card. The first results show that the proposed approach is relevant. In particular, it allows solving ambiguities for which inverting the problem is almost impossible. |
|
All publications in Ariana Research Group
|
|
2011 Activity report
