Show simple item record

hal.structure.identifierInstitut de Mathématiques de Bordeaux [IMB]
dc.contributor.authorHOUDARD, Antoine
hal.structure.identifierInstitut de Mathématiques de Bordeaux [IMB]
dc.contributor.authorLECLAIRE, Arthur
hal.structure.identifierInstitut de Mathématiques de Bordeaux [IMB]
dc.contributor.authorPAPADAKIS, Nicolas
hal.structure.identifierEquipe Image - Laboratoire GREYC - UMR6072
dc.contributor.authorRABIN, Julien
dc.date.accessioned2024-04-04T02:52:08Z
dc.date.available2024-04-04T02:52:08Z
dc.date.created2020-05-30
dc.date.conference2021-05-17
dc.identifier.urihttps://oskar-bordeaux.fr/handle/20.500.12278/192035
dc.description.abstractEnIn this paper, we propose a framework to train a generative model for texture image synthesis from a single example. To do so, we exploit the local representation of images via the space of patches, that is, square sub-images of fixed size (e.g. 4 × 4). Our main contribution is to consider optimal transport to enforce the multiscale patch distribution of generated images, which leads to two different formulations. First, a pixel-based optimization method is proposed, relying on discrete optimal transport. We show that it is related to a well-known texture optimization framework based on iterated patch nearest-neighbor projections, while avoiding some of its shortcomings. Second, in a semi-discrete setting, we exploit the differential properties of Wasserstein distances to learn a fully convolutional network for texture generation. Once estimated, this network produces realistic and arbitrarily large texture samples in real time. The two formulations result in non-convex concave problems that can be optimized efficiently with convergence properties and improved stability compared to adversarial approaches, without relying on any regularization. By directly dealing with the patch distribution of synthesized images, we also overcome limitations of state-of-the art techniques, such as patch aggregation issues that usually lead to low frequency artifacts (e.g. blurring) in traditional patch-based approaches, or statistical inconsistencies (e.g. color or patterns) in learning approaches.
dc.description.sponsorshipGeneralized Optimal Transport Models for Image processing - ANR-16-CE33-0010
dc.language.isoen
dc.title.enWasserstein Generative Models for Patch-based Texture Synthesis
dc.typeCommunication dans un congrès
dc.identifier.doi10.1007/978-3-030-75549-2_22
dc.subject.halInformatique [cs]/Apprentissage [cs.LG]
dc.subject.halInformatique [cs]/Synthèse d'image et réalité virtuelle [cs.GR]
dc.subject.halInformatique [cs]/Traitement des images
bordeaux.page269--280
bordeaux.volume12679
bordeaux.hal.laboratoriesInstitut de Mathématiques de Bordeaux (IMB) - UMR 5251*
bordeaux.institutionUniversité de Bordeaux
bordeaux.institutionBordeaux INP
bordeaux.institutionCNRS
bordeaux.conference.titleInternational Conference on Scale Space and Variational Methods in Computer Vision (SSVM'21)
bordeaux.countryFR
bordeaux.conference.cityCabourg
bordeaux.peerReviewedoui
hal.identifierhal-02824076
hal.version1
hal.invitednon
hal.proceedingsoui
hal.conference.end2021-05-19
hal.popularnon
hal.audienceInternationale
hal.origin.linkhttps://hal.archives-ouvertes.fr//hal-02824076v1
bordeaux.COinSctx_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.volume=12679&rft.spage=269--280&rft.epage=269--280&rft.au=HOUDARD,%20Antoine&LECLAIRE,%20Arthur&PAPADAKIS,%20Nicolas&RABIN,%20Julien&rft.genre=unknown


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record