Ting Peng

Ancien Doctorant, LIAMA Beijing / INRIA-Sophia

Contact :

Résumé :

Mini CV :

Dernières publications dans le projet Ariana :

Extended Phase Field Higher-Order Active Contour Models for Networks. T. Peng et I. H. Jermyn et V. Prinet et J. Zerubia. International Journal of Computer Vision, 88(1): pages 111-128, mai 2010. Mots-clés : Contour actif, Champ de Phase, Shape prior, Parameter analysis, remote sensing, Road network extraction. @ARTICLE{Peng09, author = {Peng, T. and Jermyn, I. H. and Prinet, V. and Zerubia, J.}, title = { Extended Phase Field Higher-Order Active Contour Models for Networks}, year = {2010}, month = {mai}, journal = {International Journal of Computer Vision}, volume = {88}, number = {1}, pages = { 111-128}, url = {http://www.springerlink.com/content/d3641g2227316w58/}, keyword = {Contour actif, Champ de Phase, Shape prior, Parameter analysis, remote sensing, Road network extraction} } Abstract : This paper addresses the segmentation from an image of entities that have the form of a ‘network’, i.e. the region in the image corresponding to the entity is composed of branches joining together at junctions, e.g. road or vascular networks. We present new phase field higher-order active contour (HOAC) prior models for network regions, and apply them to the segmentation of road networks from very high resolution satellite images. This is a hard problem for two reasons. First, the images are complex, with much ‘noise’ in the road region due to cars, road markings, etc., while the background is very varied, containing many features that are locally similar to roads. Second, network regions are complex to model, because they may have arbitrary topology. In particular, we address a limitation of a previous model in which network branch width was constrained to be similar to maximum network branch radius of curvature, thereby providing a poor model of networks with straight narrow branches or highly curved, wide branches. We solve this problem by introducing first an additional nonlinear nonlocal HOAC term, and then an additional linear nonlocal HOAC term to improve the computational speed. Both terms allow separate control of branch width and branch curvature, and furnish better prolongation for the same width, but the linear term has several advantages: it is more efficient, and it is able to model multiple widths simultaneously. To cope with the difficulty of parameter selection for these models, we perform a stability analysis of a long bar with a given width, and hence show how to choose the parameters of the energy functions. After adding a likelihood energy, we use both models to extract the road network quasi-automatically from pieces of a QuickBird image, and compare the results to other models in the literature. The state-of-the-art results obtained demonstrate the superiority of our new models, the importance of strong prior knowledge in general, and of the new terms in particular. New higher-order active contour models, shape priors, and multiscale analysis: their application to road network extraction from very high resolution satellite images. T. Peng. Thèse de Doctorat, Universite de Nice Sophia Antipolis, novembre 2008. Mots-clés : Contour actif d'ordre supérieur, Champ de Phase, A priori, Multiresolution, Reseaux routiers, Very high resolution. Copyright : @PHDTHESIS{Peng08d, author = {Peng, T.}, title = {New higher-order active contour models, shape priors, and multiscale analysis: their application to road network extraction from very high resolution satellite images}, year = {2008}, month = {novembre}, school = {Universite de Nice Sophia Antipolis}, pdf = {http://tel.archives-ouvertes.fr/tel-00349768/fr/}, keyword = {Contour actif d'ordre supérieur, Champ de Phase, A priori, Multiresolution, Reseaux routiers, Very high resolution} } Résumé : L'objectif de cette thèse est de développer et de valider des approches robustes d'extraction semi-automatique de réseaux routiers en zone urbaine dense à partir d'images satellitaires optiques à très haute résolution (THR). Nos modèles sont fondés sur une modélisation par champs de phase des contours actifs d'ordre supérieur (CAOS). Le probléme est difficile pour deux raisons principales : les images THR sont intrinsèquement complexes, et certaines zones des réseaux peuvent prendre une topologie arbitraire. Pour remédier à la complexité de l'information contenue dans les images THR, nous proposons une modélisation statistique multi-résolution des données ainsi qu'un modèle multi-résolution contraint a priori. Ces derniers permettent l'intégration des résultats de segmentation de résolution brute et de résolution fine. De plus, dans le cadre particulier de la mise à jour de réseaux routiers, nous présentons un modèle de forme a priori spécifique, dérivé d'une ancienne carte numérique issue d'un SIG. Ce terme spécifique a priori équilibre l'effet de la connaissance a priori générique apportée par le modèle de CAOS, qui décrit la forme géométrique générale des réseaux routiers. Cependant, le modèle classique de CAOS souffre d'une limitation importante : la largeur des branches du réseau est contrainte à d'être similaire au maximum du rayon de courbure des branches du réseau, fournissant ainsi un modèle non satisfaisant dans le cas de réseaux aux branches droites et étroites ou aux branches fortement incurvées et larges. Nous résolvons ce problème en proposant deux nouveaux modèles : l'un contenant un terme additionnel, nonlocal, non-linéaire de CAOS, et l'autre contenant un terme additionnel, nonlocal, linéaire de CAOS. Ces deux termes permettent le contrôle séparé de la largeur et de la courbure des branches, et fournissent une meilleure prolongation pour une même largeur. Le terme linéaire a plusieurs avantages : d'une part il se calcule plus efficacement, d'autre part il peut modéliser plusieurs largeurs de branche simultanément. Afin de remédier à la difficulté du choix des paramètres de ces modèles, nous analysons les conditions de stabilité pour une longue barre d'une largeur donnée décrite par ces énergies, et montrons ainsi comment choisir rigoureusement les paramètres des fonctions d'énergie. Des expériences sur des images satellitaires THR et la comparaison avec d'autres modèles démontrent la supériorité de nos modèles. Abstract : The objective of this thesis is to develop and validate robust approaches for the semi-automatic extraction of road networks in dense urban areas from very high resolution (VHR) optical satellite images. Our models are based on the recently developed higher-order active contour (HOAC) phase field framework. The problem is difficult for two main reasons: VHR images are intrinsically complex and network regions may have arbitrary topology. To tackle the complexity of the information contained in VHR images, we propose a multiresolution statistical data model and a multiresolution constrained prior model. They enable the integration of segmentation results from coarse resolution and fine resolution. Subsequently, for the particular case of road map updating, we present a specific shape prior model derived from an outdated GIS digital map. This specific prior term balances the effect of the generic prior knowledge carried by the HOAC model, which describes the geometric shape of road networks in general. However, the classical HOAC model suffers from a severe limitation: network branch width is constrained to be similar to maximum network branch radius of curvature, thereby providing a poor model of networks with straight narrow branches or highly curved, wide branches. We solve this problem by introducing two new models: one with an additional nonlinear nonlocal HOAC term, and one with an additional linear nonlocal HOAC term. Both terms allow separate control of branch width and branch curvature, and furnish better prolongation for the same width, but the linear term has several advantages: it is more efficient from a computational standpoint, and it is able to model multiple widths simultaneously. To cope with the difficulty of parameter selection of these models, we analyze the stability conditions for a long bar with a given width described by these energies, and hence show how to choose rigorously the parameters of the energy functions. Experiments on VHR satellite images and comparisons with other approaches demonstrate the superiority of our models. An extended phase field higher-order active contour model for networks and its application to road network extraction from VHR satellite images. T. Peng et I. H. Jermyn et V. Prinet et J. Zerubia. Dans Proc. European Conference on Computer Vision (ECCV), Marseille, France, octobre 2008. Mots-clés : Dense urban area, Champ de Phase, Reseaux routiers, Methodes variationnelles, Very high resolution. Copyright : @INPROCEEDINGS{Peng08c, author = {Peng, T. and Jermyn, I. H. and Prinet, V. and Zerubia, J.}, title = {An extended phase field higher-order active contour model for networks and its application to road network extraction from VHR satellite images}, year = {2008}, month = {octobre}, booktitle = {Proc. European Conference on Computer Vision (ECCV)}, address = {Marseille, France}, pdf = {http://link.springer.com/chapter/10.1007%2F978-3-540-88690-7_38}, keyword = {Dense urban area, Champ de Phase, Reseaux routiers, Methodes variationnelles, Very high resolution} } Abstract : This paper addresses the segmentation from an image of entities that have the form of a 'network', i.e. the region in the image corresponding to the entity is composed of branches joining together at junctions, e.g. road or vascular networks. We present a new phase field higher-order active contour (HOAC) prior model for network regions, and apply it to the segmentation of road networks from very high resolution satellite images. This is a hard problem for two reasons. First, the images are complex, with much 'noise' in the road region due to cars, road markings, etc., while the background is very varied, containing many features that are locally similar to roads. Second, network regions are complex to model, because they may have arbitrary topology. In particular, we address a severe limitation of a previous model in which network branch width was constrained to be similar to maximum network branch radius of curvature, thereby providing a poor model of networks with straight narrow branches or highly curved, wide branches. To solve this problem, we propose a new HOAC prior energy term, and reformulate it as a nonlocal phase field energy. We analyse the stability of the new model, and find that in addition to solving the above problem by separating the interactions between points on the same and opposite sides of a network branch, the new model permits the modelling of two widths simultaneously. The analysis also fixes some of the model parameters in terms of network width(s). After adding a likelihood energy, we use the model to extract the road network quasi-automatically from pieces of a QuickBird image, and compare the results to other models in the literature. The results demonstrate the superiority of the new model, the importance of strong prior knowledge in general, and of the new term in particular.

Liste complète des publications dans le projet Ariana