Dissertations / Theses on the topic 'Illustrations – Analyse et indexation des documents'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Illustrations – Analyse et indexation des documents.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Fournier, Jérôme. "Indexation d'images par le contenu et recherche interactive dans les bases généralistes." Cergy-Pontoise, 2002. http://biblioweb.u-cergy.fr/theses/02CERG0157.pdf.
Full textThis thesis deals with content-based image indexing and retrieval in general databases. We introduce an operational system named RETIN. From the indexing point of view, we propose an automatic processing in order to compute the image signatures. We also pay attention to dimensionality reduction and retrieval effectiveness improvement of signatures. From the retrieval point of view, we use the search-by-similarity and the relevance feedback principles in order to reduce the gap between the low-level information extracted from images and the high-level user's request. We propose a new method for the similarity function refinement and an exploration strategy for the interactive construction of a multiple request. Moreover, we introduce a long-term similarity learning technique, based on former retrieval sessions, which allows to cluster images into broad categories
Bouteldja, Nouha. "Accélération de la recherche dans les espaces de grande dimension : Application à l'indexation d'images par contenu visuel." Paris, CNAM, 2009. http://www.theses.fr/2009CNAM0628.
Full textIn this thesis we are interested in accelerating retrieval in large databases where entities are described with high dimensional vectors (or multidimensional points). Several index structures have been already proposed to accelerate retrieval but a large number of these structures suffer from the well known Curse of Dimensionality phenomenon (CoD). In the first part of this thesis we revisited the CoD phenomenon with classical indices in order to determine from which dimension these indices does not work; Our study showed that classical indices still perform well with moderate dimensions (< 30) when dealing with real data. However, needs for accelerating retrieval are not satisfied when dealing with high dimensional spaces or with large databases. The latter observations motivated our main contribution called HiPeR. HiPeR is based on a hierarchy of subspaces and indexes: it performs nearest neighbors search across spaces of different dimensions, by beginning with the lowest dimensions up to the highest ones, aiming at minimizing the effects of curse of dimensionality. Scanning the hierarchy can be done according to several scenarios that are presented for retrieval of exact as well as approximate neighbors. In this work, HiPeR has been implemented on the classical index structure VA-File, providing VA-Hierarchies. For the approximate scenario, the model of precision loss defined is probabilistic and non parametric (very little assumptions are made on the data distribution) and quality of answers can be selected by user at query time. HiPeR is evaluated for range queries on 3 real data-sets of image descriptors varying from 500,000 vectors to 4 millions. The experiments demonstrate that the hierarchy of HiPeR improves the best index structure by significantly. Reducing CPU time, whatever the scenario of retrieval. Its approximate version improves even more retrieval by saving I/O access significantly. In the last part of our thesis, we studied the particular case of multiple queries where each database entity is represented with several vectors. To accelerate retrieval with such queries different strategies were proposed to reduce I/O and CPU times. The proposed strategies were applied both to simple indices as well as to HiPeR
Landré, Jérôme. "Analyse multirésolution pour la recherche et l'indexation d'images par le contenu dans les bases de données images : application à la base d'images paléontologique Trans'Tyfipal." Dijon, 2005. http://www.theses.fr/2005DIJOS043.
Full textIn our work we propose a visual browsing method for content-based images retrieval consisting of the building of reduced increasing sizes signature vectors extracted from images descriptor vector by an expert of the domain of the images database. Signatures are used to build a fuzzy research tree with k-means algorithm (two improvements of this algorithm are given). Our demonstration software uses a web dynamic interface (PHP), image processing is performed by Intel OpenCV and IPP libraries, data is stored in a MySQL database, a Java3D interface allows to study visual behavior of images after classification. A testing protocol has been realized. Our method gives good results, in terms of computing time and quality of visual browsing results for homogeneous images databases
Julien, Charbel. "Image statistical learning using mixture models." Lyon 2, 2008. http://theses.univ-lyon2.fr/documents/lyon2/2008/julien_c.
Full textLes travaux de la thèse ont porté essentiellement sur la modélisation du contenu visuel de bas niveau des images (Couleur, Texture, etc…). La modélisation de contenu visuel est la première étape à considérer dans tout système automatique de recherche d'image par contenu, y compris les approches d'apprentissage supervisé, non-supervisé, et semi-supervisé. Dans cette thèse nous avons choisi de modéliser le contenu visuel de bas niveau, par une signature « discret distribution » ou par un modèle du mélange « GMM » au lieu des simples modèles statistiques largement utilisés dans la littérature. En utilisant ces deux types de représentation, un prototype de clustering des bases d'images a été implémenté. Ce prototype est capable d'extraire les signatures et les GMM qui représentent les images, elles sont sauvegardées pour des traitements ultérieurs y compris le clustering des images. Dans ce type de représentation les distances classiques comme la distance Euclidienne, L-2 distance, etc. Ne seront plus applicables. Des distances qui nécessitent une optimisation linéaire peuvent être utilisées pour mesurer la distance entre signatures ou GMMs, exemple : « Mallows distance » et « Earth Mover’s distance EMD ». Calculer un vecteur moyen dans le cas où on utilise des vecteurs multidimensionnels, de longueur fixe, pour représenter les images peut être relativement facile. Par contre, dans notre cas un algorithme itératif qui nécessite de nouveau une optimisation linéaire a été proposé pour apprendre un modèle, signature ou GMM, et cela en exploitant les contraintes fixées par les utilisateurs
Goëau, Hervé. "Structuration de collections d'images par apprentissage actif crédibiliste." Phd thesis, Grenoble 1, 2009. http://www.theses.fr/2009GRE10070.
Full textImage annotation is an essential task in professional archives exploitation. Archivsits must describe every image in order to make easier future retrieval tasks. The main difficulties are how to interpret the visual contents, how to bring together images whitch can be associated in same categories, and how to deal with the user's subjectivity. In this thesis, we use the principle of active learning in order to help a user who wants organize with accuracy image collections. From the visual content analysis, complementary active learning strategies are proposed to the user to help him to identify and put together images in relevant categories according to his oppinion. We choose to express this image classification problem with active learning by using the Transferable Belief Model (TBM), an elaboration on the Dempster-Shafer theory of evidence. The TBM allows the combination, the revision and the representation of the knowledge which can be extracted from the visual contents and the previously identified categories. Our method proposed in this theoritical framework gives a detailed modeling of the knowledge by representing explicitly cases of multi-labeling, while quantifying uncertainty (related to the semantic gap) and conflict induced by the analysis of the visual content in different modalities (colors, textures). A human-machine interface was developed in order to validate our approach on reference tests, personal images collections and professional photos from the National Audiovisual Institute. An evaluation was driven with professional users and showed very positive results in terms of utility, of usability and satisfaction
Belghith, Akram. "Indexation de spectres HSQC et d’images IRMf appliquée à la détection de bio-marqueurs." Thesis, Strasbourg, 2012. http://www.theses.fr/2012STRAD001/document.
Full textThe medical signal acquisition techniques are constantly evolving in recent years and providing an increasing amount of data which should be then analyzed. In this context, automatic signal processing methods are regularly proposed to assist the expert in the qualitative and quantitative analysis of these images in order to facilitate their interpretation. These methods should take into account the physics of signal acquisition, the a priori we have on the signal formation and the amount of data to analyze for a more accurate and reliable interpretation. In this thesis, we focus on the two-dimensional 2D Heteronuclear Single Quantum Coherence HSQC spectra obtained by High-Resolution Magic Angle Spinning HR-MAS NMR for biological tissue analysis and the functional Magnetic Resonance Imaging fMRI images for functional brain activities analysis. Each processed medical information will be characterized by a set of objects that we seek to extract, align, and code. The clustering of these objects by measuring their similarity will allow their classification and then the identification of biomarkers. It is this global content-based object indexing and retrieval scheme that we propose. We are interested in this thesis to properly model and integrate the a priori knowledge we have on these biological signal allowing us to propose there after appropriate methods to each indexing step and each type of signal
Ferecatu, Marin. "Recherche d'images par retour de pertinence actif utilisant des descripteurs visuels et textuels." Versailles-St Quentin en Yvelines, 2005. http://www.theses.fr/2005VERS0001.
Full textIn this thesis we explore a number of issues related to interactive querying of image databases, using both the description of the visual appearance and keyword annotations. We start by introducing the weighted color histograms, that intimately integrate color, texture and shape, and we validate their quality on multiple ground truth databases. Then, we propose several improvements for the SVM-based relevance feedback approach: first, we put forward an improved active learning selection strategy, based on a reduction of the redundancy between the images selected at every feedback round and second, we propose the use of specific kernel functions, such as the triangular kernel, that allow to obtain insensitivity to changes in the scale of the data, while keeping performance at a very good level. Finally, for image databases having keyword annotations associated with images, we propose a new keyword-based feature vector that makes use of an external ontology (WordNet) to induce a semantic generalization of the concepts corresponding to keywords. When used in combination with the visual signatures, our new feature vector can improve significantly the quality of the returned results both in query by visual example situation and with relevance feedback
Goëau, Hervé. "Structuration de collections d'images par apprentissage actif crédibiliste." Phd thesis, Université Joseph Fourier (Grenoble), 2009. http://tel.archives-ouvertes.fr/tel-00410380.
Full textJournet, Nicholas. "Analyse d’images de documents anciens : une approche texture." La Rochelle, 2006. http://www.theses.fr/2006LAROS178.
Full textMy phd thesis subject is related to the topic of old documents images indexation. The corpus of old documents has specific characteristics. The content (text and image) as well as the layout information are strongly variable. Thus, it is not possible to work on this corpus such as it usually done with contemporary documents. Indeed, the first tests which we realised on the corpus of the “Centre d’Etude de la Renaissance”, with which we work, confirmed that the traditional approaches (driven –model approaches) are not very efficient because it’s impossible to put assumptions on the physical or logical structure of the old documents. We also noted the lack of tools allowing the indexing of large old documents images databases. In this phd work, we propose a new generic method which permits characterization of the contents of old documents images. This characterization is carried out using a multirésolution study of the textures contained in the images of documents. By constructing signatures related with the frequencies and the orientations of the various parts of a page it is possible to extract, compare or to identify different kind of semantic elements (reference letters, illustrations, text, layout. . . ) without making any assumptions about the physical or logical structure of the analyzed documents. These textures information are at the origin of creation of indexing tools for large databases of old documents images
Régimbeau, Gérard. "Thematique des oeuvres plastiques contemporaines et indexation documentaire." Toulouse 2, 1996. http://www.theses.fr/1996TOU20071.
Full textDocumentary searches on contemporary art works are often based on theme. This thesis primarily sets out to explore the means and methods needed to conduct searches of this type, not so much from a strictly museographical standpoint as from a perspective wich covers all textual and iconic documents published on the subject of art works, and which make up the ressources of document centres and specialised libraries. Part 1 sets out the issues in terms of documentology, which stand at the point of intersection between information sciences and contemporary art history. Part 2 describes, analyses and evaluates the search tools contained within primary documents (indexes of books, exhibition catalogues and periodicals), as well as the indexing systems used in secondary documents (bibliographies), data bases and image banks, to ascertain what sort of thematic information that can be extracted from them. Since results are often inadequate, the theme in question needs to be reexamined and explored in relation to the field of documentation. Part 3 sets out theoritical and methodological premises, and puts forward a basis for thematic indexing of texts and reproductions in the field of contemporary art. Specific issues relating to theme in abstract and "materiological" art are also addressed. The conclusion emphasises the needs to define thematics within an interdisciplinary context wich includes iconography, cultural history and information sciences
Didier, Marie Duchemin Pierre-Yves. "Indexation, structuration et encodage des fonds iconographiques le fonds Léon Lefebvre de la Bibliothèque municipale de Lille /." [S.l.] : [s.n.], 2004. http://www.enssib.fr/bibliotheque/documents/dcb/didier.pdf.
Full textBertrand, Annick. "Compréhension et catégorisation dans une activité complexe : l'indexation de documents scientifiques." Toulouse 2, 1993. http://www.theses.fr/1993TOU20042.
Full textIndexing consists in identifying the main ideas of a document and in converging them into indexing terms chosen among a controlled vocabulary. An important variability among indexers has been shown and thereby raises two questions : what processing are made by indexers to structure and to reduce the information contained into a document and to categorize the ideas into indexing terms ? what factors step in these processing ? a situation reproducing the characteristics of the actual work has been built up. Twenty-five subjects have indexed eleven books ; they had to write the terms presenting the main ideas and the indexing terms retained from the documenting language rameau. Two classes of variables have been manipulated : the textual content (three topics and two languages) and the operators' characteristics (expertise in indexing, practice of rameau and familiarity with the scientific domain). The results reveal three indexing strategies that can vbe distinguished, on the one hand, by the nature of knowledge involved and the objectives pursued, on the other hand, by the preferential use of either bottom-up processes driven by the knowledge. The major difference depends on the level of analysis of the document : those who pratice rameau carried the level of analysisrequired by this language, the unexperienced in the language carried out a 'specific' description and the beginner indexers have chosen a general level of analysis to access quickly to the language
Blettery, Emile. "Structuring heritage iconographic collections : from automatic interlinking to semi-automatic visual validation." Electronic Thesis or Diss., Université Gustave Eiffel, 2024. http://www.theses.fr/2024UEFL2001.
Full textThis thesis explores automatic and semi-automatic structuring approaches for iconographic heritage contents collections. Indeed, exploiting such contents could prove beneficial for numerous applications. From virtual tourism to increased access for both researchers and the general public, structuring the collections would increase their accessibility and their use. However, the inherent "in silo" organization of those collections, each with their unique organization system hinders automatic structuring approaches and all subsequent applications. The computer vision community has proposed numerous automatic methods for indexing (and structuring) image collections at large scale. Exploiting the visual aspect of the contents, they are not impacted by the differences in metadata structures that mainly organize heritage collections, thus appearing as a potential solution to the problem of linking together unique data structures. However, those methods are trained on large, recent datasets, that do not reflect the visual diversity of iconographic heritage contents. This thesis aims at evaluating and exploiting those automatic methods for iconographic heritage contents structuring.To this end, this thesis proposes three distinct contributions with the common goal of ensuring a certain level of interpretability for the methods that are both evaluated and proposed. This interpretability is necessary to justify their efficiency to deal with such complex data but also to understand how to adapt them to new and different content. The first contribution of this thesis is an evaluation of existing state-of-the-art automatic content-based image retrieval (CBIR) approaches when faced with the different types of data composing iconographic heritage. This evaluation focuses first on image descriptors paramount for the image retrieval step and second, on re-ranking methods that re-order similar images after a first retrieval step based on another criterion. The most relevant approaches can then be selected for further use while the non-relevant ones provide insights for our second contribution. The second contribution consists of three novel re-ranking methods exploiting a more or less global spatial information to re-evaluate the relevance of visual similarity links created by the CBIR step. The first one exploits the first retrieved images to create an approximate 3D scene of the scene in which retrieved images are positioned to evaluate their coherence in the scene. The second one simplifies the first while extending the classical geometric verification setting by performing geometric query expansion, that is aggregating 2D geometric information from retrieved images to encode more largely the scene's geometry without the costly step of 3D scene creation. Finally, the third one exploits a more global location information, at dataset-level, to estimate the coherence of the visual similarity between images with regard to their spatial proximity. The third and final contribution is a framework for semi-automatic visual validation and manual correction of a collection's structuring. This framework exploits on one side the most suited automatic approaches evaluated or proposed earlier, and on the other side a graph-based visualization platform. We exploit several visual clues to focus the expert's manual intervention on impacting areas. We show that this guided semi-automatic approach has merits in terms of performance as it solves mistakes in the structuring that automatic methods can not, these corrections being then largely diffused throughout the structure, improving it even more globally.We hope our work will provide some first insights on automatically structuring heritage iconographic content with content-based approaches but also encourage further research on guided semi-automatic structuring of image collections
Saoudi, Aissa. "Approche spatio-temporelle pour l'indexation de documents vidéo : application à la détection de copies illégales." Paris 8, 2008. http://www.theses.fr/2008PA082944.
Full textIllegal exploitation of video content on internet became a major problem for audiovisual industry in general and for movie industry in particular. To fight against this phenomenon, content owners require new tools to help them identifying illicit file sharing, precisely and without errors. Required technologies must be at the same time relevant and robust against the various transformations which can alter video content in the copied files compared to the original ones. In this context, we propose a video search tool based on a new approach of the video spatiotemporal characterization. Initially, the approach consists in segmenting the signal in several parts, corresponding to more than the shots, thus producing temporal positions which we call "salient positions". These positions represent good pivots to ensure local relevant characterization of video contents, due to their uniform temporal distribution along the video signal and their strong persistence among the various copies. To detect the salient positions, we developed a new approach of temporal segmentation based on the analysis of SpatioTemporal Video Slices (STVS). The exploitation of the information conveyed by these slices allows us to infer the shots transitions corresponding to the salient positions. The richness of STVS in spatiotemporal information encouraged us to exploit them in the direct characterization of videos. Indeed, for each salient position, a segment of STVS is extracted then characterized using images indexing approaches. The detection of copy is achieved by similarity measurement on characteristics of STVS segments between copies and originals
Papy, Fabrice. "Hypertextualisation automatique de documents techniques." Paris 8, 1995. http://www.theses.fr/1995PA081014.
Full textAutomatic hypertextualization, an empirical process leading to hypertext, uses sequential technical documents typed from word processing software, to create dynamically the nodes and links of hypertext networks. The phase of nodes extraction uses the physical structure to delect the logical entities within documents. Referential links (especially cross-references), whose the syntax is defined by author, are extracted by means of a parser which uses a generic definition of cross-references grammar. Automatic hypertextualization produces a hypertext meta-network, where documents updating may corrupt nodes and links coherence. As relational database management systems have proved their efficiency to preserve data integrity, we propose a relational normalization of hypertextualized documents in order to manage referential links updating. Increasing of the mass of information is another outcome of the automatic creation of hypertext networks because it accentuates more disorientation problems and cognitive overhead. A solution consists of joining the hypertextualization process with an automatic indexing system, which would allow to associate each node with a set of relevant terms representing node content. So, readers will have not only structural navigation mecanisms but semantic browsing capabilities
Le, Thi-Lan. "Indexation et recherche de vidéo pour la vidéosurveillance." Nice, 2009. http://www.theses.fr/2009NICE4007.
Full textThe goal of this work is to propose a general approach for surveillance video indexing and retrieval. Based on the hypothesis that videos are pre-processed by an external video analysis module, this approach is composed of two phases : indexing phase and retrieval phase. In order to profit from the output of various video analysis modules, a general data model consisting of two main concepts, objects and events, is proposed. The indexing phase that aims at preparing data defined in the data model performs three tasks. Firstly, two new key blob detection methods in the object representation task choose for each detected object a set of key blobs associated with a weight. Secondly, the feature extraction task analyzes a number of visual and temporal features on detected objects. Finally, the indexing task computes attributes of the two concepts and stores them in the database. The retrieval phase starts with a user query and is composed of four tasks. In the formulation task, user expresses his query in a new rich query language. This query is then analyzed by the syntax parsing task. A new matching method in the matching task aims at retrieving effectively relevant results. Two proposed methods in the relevance feedback task allow to interact with the user in order to improve retrieved results. The key blob detection method has improved results of one method in the state of the art. The analysis of query language usage shows that many queries at different abstraction levels can be expressed. The matching method has proved its performance in comparison with two other methods in the state of the art. The complete approach has been validated on two videos databases coming from two projects : CARETAKER and CAVIAR. Videos for the CARETAKER project are analyzed by the VSIP platform of the Pulsar team while videos coming from CAVIAR project are manually annotated. Experiments have shown how the proposed approach is efficient and robust to retrieve the objects of interest and the complex events from surveillance videos
Hamroun, Mohamed. "Indexation et recherche par contenu visuel, sémantique et multi-niveaux des documents multimédia." Thesis, Bordeaux, 2019. http://www.theses.fr/2019BORD0372.
Full textDue to the latest technological advances, the amount of multimedia data is constantly increasing. In this context, the problem is how to effectively use this data? it is necessary to set up tools to facilitate its access and manipulation.To achieve this goal, we first propose an indexation and retrieval model for video shots (or images) by their visual content (ISE). The innovative features of ISE are as follows: (i) definition of a new descriptor "PMC" and (ii) application of the genetic algorithm (GA) to improve the retrieval (PMGA).Then, we focus on the detection of concepts in video shots (LAMIRA approach). In the same context, we propose a semi-automatic annotation method for video shots in order to improve the quality of indexation based on the GA.Then, we provide a semantic indexation method separating the data level from a conceptual level and a more abstract, contextual level. This new system also incorporates mechanisms for expanding the request and relevance feedback. To add more fluidity to the user query, the user can perform a navigation using the three levels of abstraction. Two systems called VISEN and VINAS have been set up to validate these last positions.Finally, a SIRI Framework was proposed on the basis of a multi-level indexation combining our 3 systems: ISE, VINAS and VISEN. This Framework provides a two-dimensional representation of features (high level and low level) for each image
Sayah, Salima. "Indexation d'images par moments : accès par le contenu aux documents visuels." Cachan, Ecole normale supérieure, 2007. http://www.theses.fr/2007DENS0005.
Full textIn our works, the goal was to investigate methods to define visual search keys, in order to characterize and use them in indexing and search process applied to big image database. We have to implement a fast and efficient partial indexing system. First we proposed a new interest point detector based on invariant scale and intensity Harris detector. By using this detector we obtained an efficient and repeatable salient points, Tbose points are characterized in the description step. Our shape descriptor is based on radial Chebyshev moment invariants, this descriptor is robust to geometric transformations. In order to make it more effective we used the color invariants. Afler indexing, the search step is very important, we first clustered the feature vectors by using the PDDP and KNN algorithms. Afler that we used the Gouet interest points matching algorithm, that has been efficient for big sets of points by using geometric constraints that are robust whatever the imag transformations are
Oueslati, Sami. "Modélisation pour l'hypertextualisation automatique de documents techniques : utilisation des organisateurs paralinguistiques et linguistiques." Grenoble, 2010. http://www.theses.fr/2010GRENL003.
Full textThe objective of the current study is the modeling of the hyper-textualization of technical documents. Such a modeling is based on three elements, mainly knowing the information needs of experts in a work situation, both paralinguistic and linguistic organizers, included in this kind of documents, and the types of objects that compose it. The stages of the suggested modeling are the following: - The segmentation of the content of the technical document into semantically coherent and autonomous knots. - The classification or categorization of these knots in terms of descriptive knowledge and operative knowledge. - The indexation of each knot, - The generation of inter-pertinent hypertext links. To improve the modelling of the automatic generation of knots, the categorization of knowledge carried by each knot as well as its indexation, we introduced linguistic processing. The model used is a semantic and linguistic model that was developed by Maria-Caterina MANES GALLO and Jacques ROUAULT. The elaborated modeling was tested by technician experts and was applied to several technical documents to demonstrate if it could be generalized
Ibrahim, Zein Al Abidin. "Caractérisation des structures audiovisuelles par analyse statistique des relations temporelles." Toulouse 3, 2007. http://www.theses.fr/2007TOU30069.
Full textThe aim of our work is to characterize the structure of audiovisual documents. Based on the detection of events that are characteristic of the document content or structure, we propose a representation of the temporal structure of a document, which is then used for document comparison. Existing methods by which document structures are identified usually take a priori knowledge into account. They are applied on a specific document type or on a specific document content. In our work, we have adopted another point of view and designed our method to be generic and knowledge independent so that it can be applied to any document types. Our approach is based on the analysis of the temporal relationships observed between any detected events, on the occurrence number and on a parametric representation of these relations. In order to enlarge the temporal relation analysis, we propose an algebra of relation based on this parametric representation. .
Li, Ki-Joune. "Contributions aux systèmes d’hypermédia : Modélisation et indexation des objets spatio-temporels." Lyon, INSA, 1992. http://www.theses.fr/1992ISAL0052.
Full textIn our works, we investigated two important aspects for the incorporation of spatio-temporal data into a hypermedia system: the modelling and the spatial indexing. As for the former aspect, we proposed a modelling method witch facilitates the integration of spatio-temporal data into a hypermedia system. Especially a modelling method for moving object was proposed, based on their trajectory. We have proved that spatial indexing method respecting well the spatial proximity of objects and queries, increases the hit-ratio. So a criteria, named the hierarchical variance was defined in order to quantify the spatial proximity of the spatial indexing method. By using the hierarchical variance, we have compared some important spatial indexing methods. And we have also proposed a new spatial indexing method which respect very well the spatial proximity by dynamic clustering method
Poullot, Sébastien. "Scalable Content-Based Video Copy Detection for Stream Monitoring and Video Mining." Paris, CNAM, 2009. http://www.theses.fr/2009CNAM0627.
Full textLe paysage vidéo a récemment été profondément bouleversé par de nombreuses innovations technologiques. Les méthodes et acteurs de la distribution et de la production vidéo ont notamment fortement évolués. Le nombre de canaux de diffusion télévisuels augmente continuellement et parallèlement Internet supporte de nombreux sites communautaires et blogs comportant de la vidéo. Les utilisateurs finaux sont devenus eux-mêmes auteurs et créateurs, le volume qu'ils génèrent concurrence largement celui produit par les professionnels. On peut aussi noter que les logiciels d'édition vidéo sont aujourd'hui grand public et que la personnalisation de contenus est simple et très en vogue. Les professionnels aussi réutilisent largement de vieux contenus pour faire du neuf. Une conséquence directe est l'augmentation croissante du nombre de copies diffusées et hébergées sur les réseaux. L'existence de ces copies soulèvent le problème de la protection des droits. Un ayant droit peut exprimer légitimement le besoin d'être rémunéré si un oeuvre lui appartenant est diffusé sur une chaîne. L'INA est chargé d'accomplir cette tâche en France et donc de surveiller les différents canaux pour noter ces rediffusions. Le challenge tient aux volumes à protéger et à surveiller. Le nombre d'heures numérisées est de l'ordre du demi million et le nombre de canaux de la centaine. Les documentalistes ne peuvent gérer une telle connaissance ni un tel afflux. Un pré travail automatique par ordinateur est obligatoire: un système de surveillance vidéo par le contenu. Celui-ci est chargé de lire les flux vidéos diffusés et de décider si dans ces flux apparaissent des vidéos issues de la base référence à protéger. La détection par le contenu signifie l'utilisation du signal vidéo pour faire cette reconnaissance. Les vidéos représentent de gros volumes de données, et l'exploitation du signal complet n'est pas envisageable. Par conséquent on résume les vidéos par des descripteurs, sorte de quantificateurs du signal. Le problème de la surveillance repose alors sur la recherche de descripteurs dans une base de descripteurs de référence. Ces bases contiennent des milliards de descripteurs qui sont des vecteurs de moyenne ou grande dimension (20 à quelques centaines). Un tel système pour être viable demande alors un système d'indexation des descripteurs pour effectuer des recherches rapides. Après cette recherche un processus prend la décision en utilisant les descripteurs issus de la recherche. Dans cette thèse nous présentons un nouveau schéma d'indexation, appelé Zgrid, pour faire la recherche rapide. Ce schéma permet de faire une recherche approximative. Nous l'avons amélioré par des analyses de distribution des données dans l'espace de description. Par ailleurs nous proposons un nouveau modèle des distortions subies par les descripteurs lors des processus de copies et un modèle de densité locale pour corriger la recherche, celle-ci est alors plus sélective et moins consommatrice de temps. L'utilisation croisée de ces différentes propositions permet de suivre en temps réel différé un flux vidéo et de le comparer à une base de référence de 280,000 heures de vidéo avec un simple PC. L'existence de nombreuses copies peut aussi présenter des avantages. La détection des différentes occurrences d'un même contenu peut permettre par exemple de mutualiser des annotations ou d'aider à la navigation dans les bases vidéos. Le problème prend alors une autre dimension avec une complexité quadratique: on doit rechercher l'ensemble des descripteurs d'une base sur cette même base, ce qu'on appelle communément une auto jointure par similarité. Pour réduire la complexité de cette tâche nous proposons ici un nouveau descripteur dit Glocal qui utilise des descripteurs locaux pour construire un descripteur global au niveau de l'image. Ce changement de niveau permet par ailleurs de réduire aussi la complexité du processus de décision finale. Nous proposons aussi un nouveau système d'indexation adapté à l'auto jointure par similarité et à ce descripteur. La réduction globale des temps de calculs permet de trouver les occurrences dans une base de 10,000 heures avec un simple PC mais aussi de trouver ces occurrences dans une petite base (moins de 100 heures) en 30 secondes. On peut ainsi envisager des applications « off-line » pour les administrateurs de site vidéos et « online » pour les utilisateurs
Derbas, Nadia. "Contributions à la détection de concepts et d'événements dans les documents vidéos." Thesis, Grenoble, 2014. http://www.theses.fr/2014GRENM035/document.
Full textA consequence of the rise of digital technology is that the quantity of available collections of multimedia documents is permanently and strongly increasing. The indexing of these documents became both very costly and impossible to do manually. In order to be able to analyze, classify and search multimedia documents, indexing systems have been defined. However, most of these systems suffer quality or practicability issues. Their performance is limited and depends on the data volume and data variability. Indexing systems analyze multimedia documents, looking for static concepts (bicycle, chair,...), or events (wedding, protest,...). Therefore, the variability in shapes, positions, lighting or orientation of objects hinders the process. Another aspect is that systems must be scalable. They should be able to handle big data while using reasonable amount of computing time and memory.The aim of this thesis is to improve the general performance of content-based multimedia indexing systems. Four main contributions are brought in this thesis for improving different stages of the indexing process. The first one is an "early-early fusion method" that merges different information sources in order to extract their deep correlations. This method is used for violent scenes detection in movies. The second contribution is a weakly supervised method for basic concept (objects) localization in images. This can be used afterwards as a new descriptor to help detecting complex concepts (events). The third contribution tackles the noise reduction problem on ambiguously annotated data. Two methods are proposed: a shot annotation generator, and a shot weighing method. The last contribution is a generic descriptor optimization method, based on PCA and non-linear transforms.These four contributions are tested and evaluated using reference data collections, including TRECVid and MediaEval. These contributions helped our submissions achieving very good rankings in those evaluation campaigns
Joly, Alexis. "Recherche par similarité statistique dans une grande base de signatures locales pour l'identification rapide d'extraits vidéo." La Rochelle, 2005. http://www.theses.fr/2005LAROS144.
Full textContent-based video indexing deals with techniques used to analyse and to exploit video databases without needs of any additional textual description. The work presented in this report is focused more precisely on content-based video copy detection, which is one of the emerging multimedia applications for which there is a need of a concerted effort from the database community and the computer vision community. To overcome the difficulties due to the use of very large databases, both in terms of robustness and speed, we propose a complete original and efficient strategy. The first part of this report presents the particular context of copy detection and the signatures used to describe the content of the videos. The originality of our method is that it is based both on local signatures and on a global similarity measure computed after the search in the signatures database. This similarity measure is not only a vote like other classical local approaches but it includes a registration step between candidate objects and objects retrieved by the search. The second part presents the main contribution of the thesis: A new indexing and retrieval technique belonging to the approximate similarity search techniques family. Recent works shows that trading quality for time can be widely profitable to speed-up descriptors similarity search. Whereas all other approximate techniques deal with K Nearest Neighbors search, the principle of our method is to extend the approximate paradigm to range queries. The main originality consists in determining relevant regions of the space according a theoritical model for the distortions undergone by the signatures. The method allows to determine the optimal region of the space with a high controlled probability to contain the good answer. This search paradigm is called statistical query. In practice, to simplify the access to signatures, the relevant regions are determined by using an Hilbert space filling curve and the space partition that induces. The experiments show that the technique is sublinear in database size with an assymptotically linear behavior (but only for huge databases) and that the quality performances are stable. Furthermore, they highlight that statistical queries provide a very high speed-up compared to classical exact range queries. The third part is focused on the global system assessment and the description of three applications. The experiments show that the simple theoretical distortion model is efficient enough to control the effective probability to retrieve a descriptor. They also point out that approximate similarity search is particularly profitable when using local signatures since the lost of some search results does not affect the global robustness of the detection. Furthermore, the detection results are almost invariant to strong database size growing (three orders of magnitude). The proposed approach was integrated in a difered real-time TV monitoring system which is able to control 40 000 hours of videos. The high quantity and variability of the results of this system open new data mining perspectives
Ouddan, Mohammed Amine. "Indexation et recherche des documents code source basées sur une caractérisation structuro-sémantique : application à la détection de plagiats." Université de Marne-la-Vallée, 2007. http://www.theses.fr/2007MARN0340.
Full textSource code characterization is a very complex task due the amount of similarity between computer science assignments. The various transformations that occur within a plagiarized code make the plagiarism detection more difficult. We propose a multilanguage source code retrieval system for plagiarism detection which is based on twolevel characterization approach. The first level reflects the syntactic feature of the code allowing a structural characterization of its content, and the second level relates to its functional feature allowing a semantic characterization. Our approach is based on the concept of Grammar with Actions which consists to assign significance to the parsing process in a context of characterization, and at the same time, allowing access to the structural and semantic content of the code using the grammar of its programming language. The aim idea is to translate the source code into a set of symbols sequences called characteristic sequences. In the first level of characterization we talk about structural sequences and in the second level we talk about genetic sequences. In order to quantify the similarity between characteristic sequences, we use sequence alignment techniques where the similarity rate is considered as an abstraction of the plagiarism rate between the characterized codes
Le, Roux Estelle. "Extraction d'information dans des textes libres guidée par une ontologie : faciliter l'indexation du rendu audiovisuel d'un événement par l'indexation conceptuelle de textes relatant cet événement." Paris 10, 2003. http://www.theses.fr/2003PA100027.
Full textINA needs so that its audio-visual files can be used to know the contents of its documents well. The documentalists must then view these documents and index them. In order to bring a help to the documentalists, we created an information extraction system while being useful to us of an ontology to create a semantic dictionary and syntaxical and semantic patterns. At the end, we obtain conceptual indices which could be used by the documentalists at the time of indexing. The results obtained are mitigated : the great variability of the language in the articles involves the creation of a significant number of patterns but the newspaper makes it possible to extract relevant information for INA
Saad, Motaz. "Fouille de documents et d'opinions multilingue." Electronic Thesis or Diss., Université de Lorraine, 2015. http://www.theses.fr/2015LORR0003.
Full textThe aim of this thesis is to study sentiments in comparable documents. First, we collect English, French and Arabic comparable corpora from Wikipedia and Euronews, and we align each corpus at the document level. We further gather English-Arabic news documents from local and foreign news agencies. The English documents are collected from BBC website and the Arabic documents are collected from Al-jazeera website. Second, we present a cross-lingual document similarity measure to automatically retrieve and align comparable documents. Then, we propose a cross-lingual sentiment annotation method to label source and target documents with sentiments. Finally, we use statistical measures to compare the agreement of sentiments in the source and the target pair of the comparable documents. The methods presented in this thesis are language independent and they can be applied on any language pair
Le, Lan Gaël. "Analyse en locuteurs de collections de documents multimédia." Thesis, Le Mans, 2017. http://www.theses.fr/2017LEMA1020/document.
Full textThe task of speaker diarization and linking aims at answering the question "who speaks and when?" in a collection of multimedia recordings. It is an essential step to index audiovisual contents. The task of speaker diarization and linking firstly consists in segmenting each recording in terms of speakers, before linking them across the collection. Aim is, to identify each speaker with a unique anonymous label, even for speakers appearing in multiple recordings, without any knowledge of their identity or number. The challenge of the cross-recording linking is the modeling of the within-speaker/across-recording variability: depending on the recording, a same speaker can appear in multiple acoustic conditions (in a studio, in the street...). The thesis proposes two methods to overcome this issue. Firstly, a novel neural variability compensation method is proposed, using the triplet-loss paradigm for training. Secondly, an iterative unsupervised domain adaptation process is presented, in which the system exploits the information (even inaccurate) about the data it processes, to enhance its performances on the target acoustic domain. Moreover, novel ways of analyzing the results in terms of speaker are explored, to understand the actual performance of a diarization and linking system, beyond the well-known Diarization Error Rate (DER). Systems and methods are evaluated on two TV shows of about 40 episodes, using either a global, or longitudinal linking architecture, and state of the art speaker modeling (i-vector)
Bernard, Sylvain. "Indexation et recherche dans de grandes bases d'empreintes digitales." Versailles-St Quentin en Yvelines, 2001. http://www.theses.fr/2001VERS0034.
Full textOur aim is to develop an Automated Fingerprint identification system (AFIS) for fingerprint image retrieval in Very Large Databases. We view an AFIS as a Content based image retrieval system on specific databases. As a consequence, a large domain-specific prior knowledge is incorporated into the algorithms. Indeed, our system is based on the matching of minutiae points which are the terminaisons and bifurcations of the ridge lines that constitute a fingerprint image. The efficiency of minutiae detection depends on how well the idges and valleys are extracted. The complexity of existing segmentation algorithms is often too high and quality/speed tradeoff is necessary. In our present work, we propose a fast Gabor wavelet filter bank that avoids such a tradeoff. Moreover, our multiscale approach is original and provides noise elimination whilst preserving singularities that characterize minutiae. Minutiae matching is the ultimate step used to demonstrate whether two fingerprints originate from the same finger or not. The algorithm has to be robust to rigid and not-rigid deformations, noise and occlusions. We propose a method based on a generalized Hough transform and a similarity metric that takes the geometric relationships between minutiae into account. An international benchmark proves the efficiency and robustness of the proposed method. Since minutiae matching is time consuming, matching a fingerprint with the entire database would be computationally intensive. We matching is time consuming, matching a fingerprint with the entire database would be computationally intensive. We match the query fingerprint with a subset of the database obtained by a classification step. We have developed a neural-network based classifier that organizes fingerprints from information about their global shape. The algorithm has been extensively tested on our databases providing a fast and efficient pruning
Law-To, Julien. "From genericity to distinctiveness of video content description : application to video copy detection." Versailles-St Quentin en Yvelines, 2007. http://www.theses.fr/2007VERS0025.
Full textMes travaux de thèse portent sur l’indexation et la recherche dans de grandes bases de vidéos. Partant d’une description visuelle de l’image basée sur plusieurs natures de points d’intérêt, notre approche aboutit à une représentation de plus haut niveau, associant descripteurs visuels locaux, leurs trajectoires ainsi qu’une interprétation en termes de comportement de ces descripteurs locaux au sein de la vidéo. Cette méthode permet une description fine de la vidéo tout en réduisant la redondance temporelle qui lui est intrinsèquement liée. Une application cruciale dans la gestion de patrimoines numériques est la traçabilité du catalogue vidéo. Dans ce contexte, nous proposons ViCopT, un système de détection de copie par le contenu. Une validationde sa robustesse et de sa discriminance a été réalisée sur une base de 1000h et a montrée la pertinence de nos choix. Les hautes performances de ViCopT ont été mesurées dans des évaluations comparatives tant au niveau européen qu'international
Nguyen, Dang Tuan. "Extraction d'information à partir de documents Web multilingues : une approche d'analyses structurelles." Caen, 2006. http://www.theses.fr/2006CAEN2023.
Full textMultilingual Web Document (MWD) processing has become one of the major interests of research and development in the area of information retrieval. Therefore, we observed that the structure of the multilingual resources has not been enough explored in most of the research works in this area. We consider that links structure embed crucial information for both hyperdocument retrieving and mining process. Discarding the multilingual information structures could affect the processing performance and generate various problems : i)°Redundancy : if the site proposes simultaneously translations in several languages, ii)° Noisy information: by using labels to shift from language to another, iii)° Loosing information: if the process does not consider the structure specificity of each language. In this context, we wonder to remind that each Web site is considered as a hyper-document that contains a set of Web documents (pages, screen, messages) which can be explored through the links paths. Therefore, detecting the dominant languages, in a Web Site, could be done in a different ways. The framework of this experimental research thesis is structures analysis for information extraction from a great number of heterogeneous structured or semi-structured electronic documents (essentially the Web document). It covers the following aspects : Enumerating the dominants languages, Setting-up (virtual) frontiers between those languages, enabling further processing, Recognizing the dominants languages. To experiment and validate our aim we have developed Hyperling which is a formal, language independent, system dealing with Web Documents. Hyperling proposes a Multilingual Structural Analysis approach to cluster and retrieve Web Document. Hyperling’s fundamental hypothesis is based on the notion of relation-density : The Monolingual relation density: i. E. Links between Web Documents written in the same language, The Interlingual relation density: i. E. Links between Web Documents written in different languages. In a Web document representation we can encounter a high level of monolingual relation density and low level of inter-lingual relation density. Therefore, we can consider a MWD to be represented by a set of clusters. Regarding the density level of each cluster, it may represent a dominant language. This hypothesis has been the core of Hyperling and has been experimented and approved on a real multilingual web documents (IMF, UNDP, UNFPA, UNICEF, WTO)
Ango-Obiang, Marie-France David Amos. "Proposition d'une méthodologie d'indexation d'images d'œuvres architecturales à partir de l'analyse des besoins des décideurs Propriétaire, architectes, constructeurs et administrateurs /." S. l. : Université Nancy 2, 2007. http://cyberdoc.univ-nancy2.fr/htdocs/docs_ouvert/doc289/2007NAN21013.pdf.
Full textSaad, Motaz. "Fouille de documents et d'opinions multilingue." Thesis, Université de Lorraine, 2015. http://www.theses.fr/2015LORR0003/document.
Full textThe aim of this thesis is to study sentiments in comparable documents. First, we collect English, French and Arabic comparable corpora from Wikipedia and Euronews, and we align each corpus at the document level. We further gather English-Arabic news documents from local and foreign news agencies. The English documents are collected from BBC website and the Arabic documents are collected from Al-jazeera website. Second, we present a cross-lingual document similarity measure to automatically retrieve and align comparable documents. Then, we propose a cross-lingual sentiment annotation method to label source and target documents with sentiments. Finally, we use statistical measures to compare the agreement of sentiments in the source and the target pair of the comparable documents. The methods presented in this thesis are language independent and they can be applied on any language pair
Lespinasse, Ide Lafargue Karine. "Acquisition sémantique en langue générale : la paradocumentation textuelle pour l'indexation des documents audiovisuels sur la politique." Paris 3, 2002. http://www.theses.fr/2002PA030040.
Full textIn order to respond to the documentation needs of the INA (the National Broadcasting Institute), which indexes French television program archives, we propose to use "paradocumentation" in computer-aided treatments. The so-called "paradocumentation" is comprised of all of the textual or photographic documents produced in the making of the audiovisual document. As long as the "peritexts" exist in an electronic format, they can be structured in the form of corpora of which the content is closely linked to the audiovisual program and can therefore enrich its description. The scientific goal is to exploit the results obtained from the treatments in a general language with an approach historically developed for sublanguages (methods, tools). There is a pragmatic reason for this choice (no other tools are available) as well as a theoretical one : general language resists automation as showed by Zellig Harris. We have conducted several experiments on the semantic acquisition of different peritexts taken from political programs : indexing files, transcriptions from soundtracks, press articles. .
Trichili, Hanène. "Élaboration d'une nouvelle approche de tatouage pour l'indexation des images médicales." Télécom Bretagne, 2006. http://www.theses.fr/2006TELB0006.
Full textPham, Nguyen-Khang. "Analyse factorielle des correspondances pour l'indexation et la recherche d'information dans une grande base de données d'images." Rennes 1, 2009. http://www.theses.fr/2009REN1S067.
Full textWith the development of the digital world, the number of images stored in databases has significantly increased. Image indexing and information retrieval in image databases are more complicated than in the case of textual documents. Indexing methods already used in textual data analysis are proposed to process images. To transfer the results of the textual data analysis to images, new features are required: visual words and images are considered as documents. We are interested in the problem of indexing and information retrieval in a large database of images using data analysis methods and, more specifically, using Factorial Correspondence Analysis (FCA). First, we propose to use relevant indicators of FCA to speed up the retrieval step after adapting it to images. Next, we study the large scale retrieval with FCA. To this end, we propose an incremental FCA algorithm to deal with large contingency tables, and its parallelization on Graphics Processing Units (GPUs). We also develop a parallel version of our search algorithm using relevant indicators of FCA on GPUs. After that, we combine the use of FCA with other methods such as the Contextual Dissimilarity Measure and random forests in order to improve the retrieval quality. Finally, we present a visualization environment, CAViz, which allows us to display the results
Ouwayed, Nazih. "Segmentation en lignes de documents anciens : applications aux documents arabes." Thesis, Nancy 2, 2010. http://www.theses.fr/2010NAN23001/document.
Full textThe indexing of handwritten scanned documents poses the problem of lines segmentation, if it fails, disabling the following steps of words extraction and recognition. In addition, the ancient Arabic documents contain annotations in the margins, often composed of lines obliquely oriented. The detection of these lines is important as the rest and is a major challenge for the indexing of these documents. Thus, the segmentation described in this thesis involves the extraction of multi-oriented lines. For this problem, the bibliography has only rudimentary techniques based essentially on the projection of the document image along one direction, which be failed in the case of multi-oriented documents. Given this lack, we have proposed an adaptive approach that first locates the different orientation zones, then based on each local orientation to extract the lines. During my thesis, i particularly invested on the following points : - Applying an automatic paving using the active contour model (snake). - Preparation the signal of the projection profile by removing all pixels that are not needed in the orientation estimation. Then, implementation of all energy distributions of Cohen's class on the projection profile to find the best distribution that gives the orientation. - Applying some extension rules to find the oriented zones. - Extraction of lines by using an connected components follow-up algorithm. - Separation of overlapped and touched lines using the morphology of Arabic terminal letters
Thlithi, Marwa. "Segmentation et regroupement en chanteurs : application aux enregistrements ethnomusicologiques." Thesis, Toulouse 3, 2016. http://www.theses.fr/2016TOU30059/document.
Full textThis work was done in the context of the ANR CONTINT DIADEMS project on indexing ethno-musicological audio recordings. The data that we are studying are provided by the Musée de l'Homme, Paris, within the context of this project. The work performed in this thesis consists of developing automatic structuring methods of musical and ethno-musicological documents based on the persons. This thesis touchs on an unexplored subject in our knowledge of the segmentation and clustering in singers of musical recordings. We propose a complete system in this subject that we called singer diarization by analogy with speaker diarization system on speech context. Indeed, this system is inspired from existing studies performed in speaker diarization and is designed to work on studio music recordings as well as on recordings with a variable sound quality (done outdoors). The first step of this system is the segmentation in singer turns which consists of segmenting musical recordings into segments "acoustically homogeneous" by singer group. The second step is the clustering which consists of labelling all segments produced by the same group of singers with a unique identifier. Our first contribution involved the definition of the term " singer turns " and the proposal of rules for manual annotation in singer turns segments. The second consisted in the proposal of a feature extraction method for the characterization of singer voices by implementing a method to select the frequency coefficients, which are the most relevant, based on the variance of these coefficients. The third is the implementation of a dynamic segmentation algorithm adapted to the singing context by using the Bayesian Information Criterion (BIC). The fourth is the proposal of a method, called DCAP, to take a posteriori decisions in order to avoid the variability problem of the BIC penalty parameter. Indeed, a priori choice of an optimal value for this parameter is not possible. This led us to perform a majority voting on a several segmentations obtained with different values of this parameter. A gain of about 8% and 15% is obtained on our two corpora with this method compared to the results found with a standard value of the penalty parameter. The fifth is the adaptation of our DCAP method in order to perform singer clustering step
Ouwayed, Nazih. "Segmentation en lignes de documents anciens : applications aux documents arabes." Electronic Thesis or Diss., Nancy 2, 2010. http://www.theses.fr/2010NAN23001.
Full textThe indexing of handwritten scanned documents poses the problem of lines segmentation, if it fails, disabling the following steps of words extraction and recognition. In addition, the ancient Arabic documents contain annotations in the margins, often composed of lines obliquely oriented. The detection of these lines is important as the rest and is a major challenge for the indexing of these documents. Thus, the segmentation described in this thesis involves the extraction of multi-oriented lines. For this problem, the bibliography has only rudimentary techniques based essentially on the projection of the document image along one direction, which be failed in the case of multi-oriented documents. Given this lack, we have proposed an adaptive approach that first locates the different orientation zones, then based on each local orientation to extract the lines. During my thesis, i particularly invested on the following points : - Applying an automatic paving using the active contour model (snake). - Preparation the signal of the projection profile by removing all pixels that are not needed in the orientation estimation. Then, implementation of all energy distributions of Cohen's class on the projection profile to find the best distribution that gives the orientation. - Applying some extension rules to find the oriented zones. - Extraction of lines by using an connected components follow-up algorithm. - Separation of overlapped and touched lines using the morphology of Arabic terminal letters
Berrut, Catherine. "Une méthode d'indexation fondée sur l'analyse sémantique de documents spécialisés : le prototype RIME et son application à un corpus médical." Phd thesis, Grenoble 1, 1988. http://tel.archives-ouvertes.fr/tel-00330027.
Full textPeña, Saldarriaga Sebastián. "Approches textuelles pour la catégorisation et la recherche de documents manuscrits en-ligne." Phd thesis, Université de Nantes, 2010. http://tel.archives-ouvertes.fr/tel-00483684.
Full textBouzayani, Abdessalem. "Extension automatique de l'annotation d'images pour la recherche et la classification." Electronic Thesis or Diss., Université de Lorraine, 2018. http://www.theses.fr/2018LORR0045.
Full textThis thesis deals the problem of image annotation extension. Indeed, the fast growth of available visual contents has led a need for indexing and searching of multimedia information methods. Image annotation allows indexing and searching in a large collection of images in an easy and fast way. We wish, from partially manually annotated images databases, complete automatically the annotation of these sets, in order to make methods of research and / or classification of images more efficient. For automatic image annotation extension, we use probabilistic graphical models. The proposed model is based on a mixture of multinomial distributions and mixtures of Gaussian where we have combined visual and textual characteristics. To reduce the cost of manual annotation and improve the quality of the annotation obtained, we have incorporated user feedback into our model. User feedback was done using learning in learning, incremental learning and active learning. To reduce the semantic gap problem and to enrich the image annotation, we use a semantic hierarchy by modeling many semantic relationships between keywords. We present a semi-automatic method to build a semantic hierarchy from a set of keywords. After building the hierarchy, we integrate it into our image annotation model. The model obtained with this hierarchy is a mixture of Bernoulli distributions and Gaussian mixtures
Pham, The Anh. "Détection robuste de jonctions et points d'intérêt dans les images et indexation rapide de caractéristiques dans un espace de grande dimension." Thesis, Tours, 2013. http://www.theses.fr/2013TOUR4023/document.
Full textLocal features are of central importance to deal with many different problems in image analysis and understanding including image registration, object detection and recognition, image retrieval, etc. Over the years, many local detectors have been presented to detect such features. Such a local detector usually works well for some particular applications but not all. Taking an application of image retrieval in large database as an example, an efficient method for detecting binary features should be preferred to other real-valued feature detection methods. The reason is easily seen: it is expected to have a reasonable precision of retrieval results but the time response must be as fast as possible. Generally, local features are used in combination with an indexing scheme. This is highly needed for the case where the dataset is composed of billions of data points, each of which is in a high-dimensional feature vector space
Bursuc, Andrei. "Indexation et recherche de contenus par objet visuel." Phd thesis, Ecole Nationale Supérieure des Mines de Paris, 2012. http://pastel.archives-ouvertes.fr/pastel-00873966.
Full textFiorini, Nicolas. "Semantic similarities at the core of generic indexing and clustering approaches." Thesis, Montpellier, 2015. http://www.theses.fr/2015MONTS178/document.
Full textIn order to improve the exploitation of even growing number of electronic documents, Artificial Intelligence has dedicated a lot of effort to the creation and use of systems grounded on knowledge bases. In particular in the information retrieval field, such semantic approaches have proved their efficiency.Therefore, indexing documents is a necessary task. It consists of associating them with sets of terms that describe their content. These terms can be keywords but also concepts from an ontology, in which case the annotation is said to be semantic and benefit from the inherent properties of ontologies which are the absence of ambiguities.Most approaches designed to annotate documents have to parse them and extract concepts from this parsing. This underlines the dependance of such approaches to the type of documents, since parsing requires dedicated algorithms.On the other hand, approaches that solely rely on semantic annotations can ignore the document type, enabling the creation of generic processes. This thesis capitalizes on genericity to build novel systems and compare them to state-of-the-art approaches. To this end, we rely on semantic annotations coupled with semantic similarity measures. Of course, such generic approaches can then be enriched with type-specific ones, which would further increase the quality of the results.First of all, this work explores the relevance of this paradigm for indexing documents. The idea is to rely on already annotated close documents to annotate a target document. We define a heuristic algorithm for this purpose that uses the semantic annotations of these close documents and semantic similarities to provide a generic indexing method. This results in USI (User-oriented Semantic Indexer) that we show to perform as well as best current systems while being faster.Second of all, this idea is extended to another task, clustering. Clustering is a very common and ancient process that is very useful for finding documents or understanding a set of documents. We propose a hierarchical clustering algorithm that reuses the same components of classical methods to provide a novel one applicable to any kind of documents. Another benefit of this approach is that when documents are grouped together, the group can be annotated by using our indexing algorithm. Therefore, the result is not only a hierarchy of clusters containing documents as clusters are actually described by concepts as well. This helps a lot to better understand the results of the clustering.This thesis shows that apart from enhancing classical approaches, building conceptual approaches allows us to abstract them and provide a generic framework. Yet, while bringing easy-to-set-up methods – as long as documents are semantically annotated –, genericity does not prevent us from mixing these methods with type-specific ones, in other words creating hybrid methods
Bernard, Michel. "Élaboration d'un thésaurus pour l'indexation thématique d'oeuvres littéraires." Paris 3, 1992. http://www.theses.fr/1992PA030129.
Full textThis thesis gives an account of the building up of a thesaurus of literary terms for the indexing of the french literary works listed in the literature data bank, worked out at the Paris III-Sorbonne nouvelle university. The aim has been to give a fair representation of the contents of literary texts in the french language, using the received terminology. The thesaurus, in accordance with the norms of information retrieval, distributes the vocabulary into allowed and not allowed terms. A first version was used for the indexation, by students, of 1500 works. The results of that first phase were then used in their turn for the improvement of the thesaurus, the themes’ list and its structure. It now has the form of a non hierarchical semantic network. Besides, a number of tests in automatic indexation have shown that computer-aided thematic indexation, based on abstracts, is worth considering. Statistical studies prefigure the kind of literary studies that a thematic data base would allow. Such studies would bear on the correlations between themes and movements, epochs, genres, etc. .
Tarafdar, Arundhati. "Wordspotting from multilingual and stylistic documents." Thesis, Tours, 2017. http://www.theses.fr/2017TOUR4022/document.
Full textWord spotting in graphical documents is a very challenging task. To address such scenarios this thesis deals with developing a word spotting system dedicated to geographical documents with Bangla and English (Roman) scripts. In the proposed system, at first, text-graphics layers are separated using filtering, clustering and self-reinforcement through classifier. Additionally, instead of using binary decision we have used probabilistic measurement to represent the text components. Subsequently, in the text layer, character segmentation approach is applied using water-reservoir based method to extract individual character from the document. Then recognition of these isolated characters is done using rotation invariant feature, coupled with SVM classifier. Well recognized characters are then grouped based on their sizes. Initial spotting is started to find a query word among those groups of characters. In case if the system could spot a word partially due to any noise, SIFT is applied to identify missing portion of that partial spotting. Experimental results on Roman and Bangla scripts document images show that the method is feasible to spot a location in text labeled graphical documents. Experiments are done on an annotated dataset which was developed for this work. We have made this annotated dataset available publicly for other researchers
Loiseau, Mathieu. "Elaboration d'un modèle pour une base de textes indexée pédagogiquement pour l'enseignement des langues." Grenoble 3, 2009. https://tel.archives-ouvertes.fr/tel-00440460v3.
Full textThis PhD thesis deals with the notion of pedagogical indexation and tackles it from the point of view of searching for and selecting texts for language teaching. This particular problem is set in the field of Computer Assisted Language Learning (CALL) and of the potential contribution of Natural Language Processing (NLP) to this discipline, before being considered within the scope of elements more directly relevant to language didactics, in order to propose an empirical approach. The latter is then justified by the inadequacy of current description standards for pedagogical resources where modeling of raw objects in a consistent fashion is concerned. This is particularly true for texts in the context of language learning. The thesis subsequently revolves around two questionnaires the aim of which is to provide insight into language teachers' declared practices regarding searching for and selecting texts in the context of class planning. The first questionnaire provides data to formalize the notion of pedagogical context, which is later considered through some of its components thanks to the second questionnaire. Finally, these first formalization drafts provide foundations for the definition of a model aiming at taking into account the contextuality of the properties said to be pedagogical, which is inherent to raw resources. Finally, possible leads for implementing this model are suggested through the description of a computerized system
Catteau, Olivier. "Le cycle de vie de l'objet pédagogique et de ses métadonnées." Phd thesis, Toulouse 3, 2008. http://tel.archives-ouvertes.fr/tel-00344701.
Full text- une représentation des relations qui utilise des techniques de visualisation de l'information pour rechercher des OP et fournir aux utilisateurs une vue générale de la progression de la production ;
- un service de gestion des commentaires qui offre aux enseignants et aux apprenants l'opportunité d'exploiter, d'exprimer et de partager des évaluations et des suggestions d'utilisation d'OP dans le système adéquat au moment où elles sont pertinentes ;
- un service de gestion des évolutions qui sensibilise les enseignants et les responsables de cours aux divergences qui apparaissent entre les OP importés dans les plates-formes pédagogiques et leurs évolutions et dépendances qui sont stockées dans des viviers hétérogènes.
Cette approche, basée sur des objets pédagogiques de forte granularité, a été expérimentée au sein du campus numérique International E-Mi@ge.
Kieu, Van Cuong. "Modèle de dégradation d’images de documents anciens pour la génération de données semi-synthétiques." Thesis, La Rochelle, 2014. http://www.theses.fr/2014LAROS029/document.
Full textIn the last two decades, the increase in document image digitization projects results in scientific effervescence for conceiving document image processing and analysis algorithms (handwritten recognition, structure document analysis, spotting and indexing / retrieval graphical elements, etc.). A number of successful algorithms are based on learning (supervised, semi-supervised or unsupervised). In order to train such algorithms and to compare their performances, the scientific community on document image analysis needs many publicly available annotated document image databases. Their contents must be exhaustive enough to be representative of the possible variations in the documents to process / analyze. To create real document image databases, one needs an automatic or a manual annotation process. The performance of an automatic annotation process is proportional to the quality and completeness of these databases, and therefore annotation remains largely manual. Regarding the manual process, it is complicated, subjective, and tedious. To overcome such difficulties, several crowd-sourcing initiatives have been proposed, and some of them being modelled as a game to be more attractive. Such processes reduce significantly the price andsubjectivity of annotation, but difficulties still exist. For example, transcription and textline alignment have to be carried out manually. Since the 1990s, alternative document image generation approaches have been proposed including in generating semi-synthetic document images mimicking real ones. Semi-synthetic document image generation allows creating rapidly and cheaply benchmarking databases for evaluating the performances and trainingdocument processing and analysis algorithms. In the context of the project DIGIDOC (Document Image diGitisation with Interactive DescriptiOn Capability) funded by ANR (Agence Nationale de la Recherche), we focus on semi-synthetic document image generation adapted to ancient documents. First, we investigate new degradation models or adapt existing degradation models to ancient documents such as bleed-through model, distortion model, character degradation model, etc. Second, we apply such degradation models to generate semi-synthetic document image databases for performance evaluation (e.g the competition ICDAR2013, GREC2013) or for performance improvement (by re-training a handwritten recognition system, a segmentation system, and a binarisation system). This research work raises many collaboration opportunities with other researchers to share our experimental results with our scientific community. This collaborative work also helps us to validate our degradation models and to prove the efficiency of semi-synthetic document images for performance evaluation and re-training