Tesis sobre el tema "Recherche d'images par contenu visuel"
Crea una cita precisa en los estilos APA, MLA, Chicago, Harvard y otros
Consulte los 50 mejores tesis para su investigación sobre el tema "Recherche d'images par contenu visuel".
Junto a cada fuente en la lista de referencias hay un botón "Agregar a la bibliografía". Pulsa este botón, y generaremos automáticamente la referencia bibliográfica para la obra elegida en el estilo de cita que necesites: APA, MLA, Harvard, Vancouver, Chicago, etc.
También puede descargar el texto completo de la publicación académica en formato pdf y leer en línea su resumen siempre que esté disponible en los metadatos.
Explore tesis sobre una amplia variedad de disciplinas y organice su bibliografía correctamente.
Hoàng, Nguyen Vu. "Prise en compte des relations spatiales contextuelles dans la recherche d'images par contenu visuel". Paris 9, 2011. http://basepub.dauphine.fr/xmlui/handle/123456789/8202.
Texto completoThis thesis is focused on the study of methods for image retrieval by visual content in collection of heterogeneous contents. We are interested in the description of spatial relationships between the entities present in the images that can be symbolic objects or visual primitives such as interest points. The first part of this thesis is dedicated to a state of the art on the description of spatial relationship techniques. As a result of this study, we propose the approach Δ-TSR, our first contribution, which allows similarity search based on visual content by using the triangular relationships between entities in images. In our experiments, the entities are local visual features based on salient points represented in a bag of features model. This approach improves not only the quality of the images retrieval but also the execution time in comparison with other approaches in the literature. The second part is dedicated to the study of the image context. The spatial relationships between entities in an image allow creating the global description of the image that we call the image context. Taking into account the contextual spatial relationships in the similarity search of images can allow improving the retrieval quality by limiting false alarms. We defined the context of image as the presence of entity categories and their spatial relationships in the image. We studied the relationships between different entity categories on LabelMe, a state of the art of symbolic images databases of heterogeneous content. This statistical study, our second contribution, allows creating a cartography of their spatial relationships. It can be integrated in a graph-based model of the contextual relationships, our third contribution. This graph describes the general knowledge of every entity categories. Spatial reasoning on this knowledge graph can help improving tasks of image processing such as detection and localization of an entity category by using the presence of another reference. Further, this model can be applied to represent the context of an image. The similarity search based on context can be achieved by comparing the graphs, then, contextual similarity between two images is evaluated by the similarity between their graphs. This work was evaluated on the symbolic image database of LabelMe. The experiments showed its relevance for image retrieval by context
Michaud, Dorian. "Indexation bio-inspirée pour la recherche d'images par similarité". Thesis, Poitiers, 2018. http://www.theses.fr/2018POIT2288/document.
Texto completoImage Retrieval is still a very active field of image processing as the number of available image datasets continuously increases.One of the principal objectives of Content-Based Image Retrieval (CBIR) is to return the most similar images to a given query with respect to their visual content.Our work fits in a very specific application context: indexing small expert image datasets, with no prior knowledge on the images. Because of the image complexity, one of our contributions is the choice of effective descriptors from literature placed in direct competition.Two strategies are used to combine features: a psycho-visual one and a statistical one.In this context, we propose an unsupervised and adaptive framework based on the well-known bags of visual words and phrases models that select relevant visual descriptors for each keypoint to construct a more discriminative image representation.Experiments show the interest of using this this type of methodologies during a time when convolutional neural networks are ubiquitous.We also propose a study about semi interactive retrieval to improve the accuracy of CBIR systems by using the knowledge of the expert users
Fauqueur, Julien. "Contributions pour la Recherche d'Images par Composantes Visuelles". Phd thesis, Université de Versailles-Saint Quentin en Yvelines, 2003. http://tel.archives-ouvertes.fr/tel-00007090.
Texto completoUn système de recherche d'information visuelle doit permettre à l'utilisateur de désigner d'une manière explicite la cible visuelle qu'il recherche se rapportant aux différentes composantes de l'image. Notre objectif au cours de ce travail a été de réfléchir à comment définir des clés de recherche visuelle permettant à l'utilisateur d'exprimer cette cible visuelle, de concevoir et d'implémenter efficacement les méthodes correspondantes.
Les contributions originales de cette thèse portent sur de nouvelles approches permettant de retrouver des images à partir de leurs différentes composantes visuelles selon deux paradigmes de recherche distincts.
Le premier paradigme est celui de la recherche par région exemple. Il consiste à retrouver les images comportant une partie d'image similaire à une partie visuelle requête. Pour ce paradigme, nous avons mis au point une approche de segmentation grossière en régions et de description fine de ces régions ensuite. Les régions grossières des images de la base, extraites par notre nouvel algorithme de segmentation non supervisée, représentent les composantes visuellement saillantes de chaque image. Cette décomposition permet à l'utilisateur de désigner séparément une région d'intérêt pour sa requête. La recherche de régions similaires dans les images de la base repose sur un nouveau descripteur de régions (ADCS). Il offre une caractérisation fine, compacte et adaptative de l'apparence photométrique des régions, afin de tenir compte de la spécificité d'une base de descripteurs de régions. Dans cette nouvelle approche, la segmentation est rapide et les régions extraites sont intuitives pour l'utilisateur. La finesse de description des régions améliore la similarité des régions retournées par rapport aux descripteurs existants, compte tenu de la fidélité accrue au contenu des régions.
Notre seconde contribution porte sur l'élaboration d'un nouveau paradigme de recherche d'images par composition logique de catégories de régions. Ce paradigme présente l'avantage d'apporter une solution au problème de la page zéro. Il permet d'atteindre les images, quand elles existent dans la base, qui se rapprochent de la représentation mentale de la cible visuelle de l'utilisateur. Ainsi aucune image ou région exemple n'est nécessaire au moment de la formulation de la requête. Ce paradigme repose sur la génération non-supervisée d'un thésaurus photométrique constitué par le résumé visuel des régions de la base. Pour formuler sa requête, l'utilisateur accède directement à ce résumé en disposant d'opérateurs de composition logique de ces différentes parties visuelles. Il est à noter qu'un item visuel dans ce résumé est un représentant d'une classe photométrique de régions. Les requêtes logiques sur le contenu des images s'apparentent à celles en recherche de texte. L'originalité de ce paradigme ouvre des perspectives riches pour de futurs travaux en recherche d'information visuelle.
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.
Texto completoIn 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
Le, Huu Ton. "Improving image representation using image saliency and information gain". Thesis, Poitiers, 2015. http://www.theses.fr/2015POIT2287/document.
Texto completoNowadays, along with the development of multimedia technology, content based image retrieval (CBIR) has become an interesting and active research topic with an increasing number of application domains: image indexing and retrieval, face recognition, event detection, hand writing scanning, objects detection and tracking, image classification, landmark detection... One of the most popular models in CBIR is Bag of Visual Words (BoVW) which is inspired by Bag of Words model from Information Retrieval field. In BoVW model, images are represented by histograms of visual words from a visual vocabulary. By comparing the images signatures, we can tell the difference between images. Image representation plays an important role in a CBIR system as it determines the precision of the retrieval results.In this thesis, image representation problem is addressed. Our first contribution is to propose a new framework for visual vocabulary construction using information gain (IG) values. The IG values are computed by a weighting scheme combined with a visual attention model. Secondly, we propose to use visual attention model to improve the performance of the proposed BoVW model. This contribution addresses the importance of saliency key-points in the images by a study on the saliency of local feature detectors. Inspired from the results from this study, we use saliency as a weighting or an additional histogram for image representation.The last contribution of this thesis to CBIR shows how our framework enhances the BoVP model. Finally, a query expansion technique is employed to increase the retrieval scores on both BoVW and BoVP models
Leveau, Valentin. "Représentations d'images basées sur un principe de voisins partagés pour la classification fine". Thesis, Montpellier, 2016. http://www.theses.fr/2016MONTT257/document.
Texto completoThis thesis focuses on the issue of fine-grained classification which is a particular classification task where classes may be visually distinguishable only from subtle localized details and where background often acts as a source of noise. This work is mainly motivated by the need to devise finer image representations to address such fine-grained classification tasks by encoding enough localized discriminant information such as spatial arrangement of local features.To this aim, the main research line we investigate in this work relies on spatially localized similarities between images computed thanks to efficient approximate nearest neighbor search techniques and localized parametric geometry. The main originality of our approach is to embed such spatially consistent localized similarities into a high-dimensional global image representation that preserves the spatial arrangement of the fine-grained visual patterns (contrary to traditional encoding methods such as BoW, Fisher or VLAD Vectors). In a nutshell, this is done by considering all raw patches of the training set as a large visual vocabulary and by explicitly encoding their similarity to the query image. In more details:The first contribution proposed in this work is a classification scheme based on a spatially consistent k-nn classifier that relies on pooling similarity scores between local features of the query and those of the similar retrieved images in the vocabulary set. As this set can be composed of a lot of local descriptors, we propose to scale up our approach by using approximate k-nearest neighbors search methods. Then, the main contribution of this work is a new aggregation-based explicit embedding derived from a newly introduced match kernel based on shared nearest neighbors of localized feature vectors combined with local geometric constraints. The originality of this new similarity-based representation space is that it directly integrates spatially localized geometric information in the aggregation process.Finally, as a third contribution, we proposed a strategy to drastically reduce, by up to two orders of magnitude, the high-dimensionality of the previously introduced over-complete image representation while still providing competitive image classification performance.We validated our approaches by conducting a series of experiments on several classification tasks involving rigid objects such as FlickrsLogos32 or Vehicles29 but also on tasks involving finer visual knowledge such as FGVC-Aircrafts, Oxford-Flower102 or CUB-Birds200. We also demonstrated significant results on fine-grained audio classification tasks such as the LifeCLEF 2015 bird species identification challenge by proposing a temporal extension of our image representation. Finally, we notably showed that our dimensionality reduction technique used on top of our representation resulted in highly interpretable visual vocabulary composed of the most representative image regions for different visual concepts of the training base
Landre, 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". Phd thesis, Université de Bourgogne, 2005. http://tel.archives-ouvertes.fr/tel-00079897.
Texto completo1) La taille du vecteur descripteur (n>100) rend les calculs de distance sensibles à la malédiction de la dimension,
2) La présence d'attributs de nature différente dans le vecteur descripteur ne facilite pas la classification,
3) La classification ne s'adapte pas (en général) au contexte de recherche de l'utilisateur.
Nous proposons dans ce travail une méthode basée sur la construction de hiérarchies de signatures de tailles réduites croissantes qui permettent de prendre en compte le contexte de recherche de l'utilisateur. Notre méthode tend à imiter le comportement de la vision humaine.
Le vecteur descripteur contient des attributs issus de l'analyse multirésolution des images. Ces attributs sont organisés par un expert du domaine de la base d'images en plusieurs hiérarchies de quatre vecteur signature de taille réduite croissante (respectivement 4, 6, 8 et 10 attributs). Ces signatures sont utilisées pour construire un arbre de recherche flou grâce à l'algorithme des nuées dynamiques (dont deux améliorations sont proposées). Les utilisateurs en ligne choisissent une hiérarchie de signature parmi celles proposées par l'expert en fonction de leur contexte de recherche.
Un logiciel de démonstration a été développé. Il utilise une interface web dynamique (PHP), les traitements d'images (optimisés) sont réalisés grâce aux librairies Intel IPP et OpenCV, le stockage et l'indexation sont réalisés par une base de données MySQL, une interface de visualisation 3D (Java3D) permet de se rendre compte de la répartition des images dans la classification.
Un protocole de tests psycho-visuels a été réalisé. Les résultats sur la base paléontologique Trans'Tyfipal sont présentés et offrent des réponses pertinentes selon le contexte de recherche. La méthode donne de bons résultats, tant en temps de calcul qu'en pertinence des images résultats lors de la navigation dans les bases d'images homogènes.
Gbehounou, Syntyche. "Indexation de bases d'images : évaluation de l'impact émotionnel". Thesis, Poitiers, 2014. http://www.theses.fr/2014POIT2295/document.
Texto completoThe goal of this work is to propose an efficient approach for emotional impact recognition based on CBIR techniques (descriptors, image representation). The main idea relies in classifying images according to their emotion which can be "Negative", "Neutral" or "Positive". Emotion is related to the image content and also to the personnal feelings. To achieve our goal we firstly need a correct assessed image database. Our first contribution is about this aspect. We proposed a set of 350 diversifed images rated by people around the world. Added to our choice to use CBIR methods, we studied the impact of visual saliency for the subjective evaluations and interest region segmentation for classification. The results are really interesting and prove that the CBIR methods are usefull for emotion recognition. The chosen desciptors are complementary and their performance are consistent on the database we have built and on IAPS, reference database for the analysis of the image emotional impact
Niaz, Usman. "Amélioration de la détection des concepts dans les vidéos en coupant de plus grandes tranches du monde visuel". Electronic Thesis or Diss., Paris, ENST, 2014. http://www.theses.fr/2014ENST0040.
Texto completoVisual material comprising images and videos is growing ever so rapidly over the internet and in our personal collections. This necessitates automatic understanding of the visual content which calls for the conception of intelligent methods to correctly index, search and retrieve images and videos. This thesis aims at improving the automatic detection of concepts in the internet videos by exploring all the available information and putting the most beneficial out of it to good use. Our contributions address various levels of the concept detection framework and can be divided into three main parts. The first part improves the Bag of Words (BOW) video representation model by proposing a novel BOW construction mechanism using concept labels and by including a refinement to the BOW signature based on the distribution of its elements. We then devise methods to incorporate knowledge from similar and dissimilar entities to build improved recognition models in the second part. Here we look at the potential information that the concepts share and build models for meta-concepts from which concept specific results are derived. This improves recognition for concepts lacking labeled examples. Lastly we contrive certain semi-supervised learning methods to get the best of the substantial amount of unlabeled data. We propose techniques to improve the semi-supervised cotraining algorithm with optimal view selection
Niaz, Usman. "Amélioration de la détection des concepts dans les vidéos en coupant de plus grandes tranches du monde visuel". Thesis, Paris, ENST, 2014. http://www.theses.fr/2014ENST0040/document.
Texto completoVisual material comprising images and videos is growing ever so rapidly over the internet and in our personal collections. This necessitates automatic understanding of the visual content which calls for the conception of intelligent methods to correctly index, search and retrieve images and videos. This thesis aims at improving the automatic detection of concepts in the internet videos by exploring all the available information and putting the most beneficial out of it to good use. Our contributions address various levels of the concept detection framework and can be divided into three main parts. The first part improves the Bag of Words (BOW) video representation model by proposing a novel BOW construction mechanism using concept labels and by including a refinement to the BOW signature based on the distribution of its elements. We then devise methods to incorporate knowledge from similar and dissimilar entities to build improved recognition models in the second part. Here we look at the potential information that the concepts share and build models for meta-concepts from which concept specific results are derived. This improves recognition for concepts lacking labeled examples. Lastly we contrive certain semi-supervised learning methods to get the best of the substantial amount of unlabeled data. We propose techniques to improve the semi-supervised cotraining algorithm with optimal view selection
Risser-Maroix, Olivier. "Similarité visuelle et apprentissage de représentations". Electronic Thesis or Diss., Université Paris Cité, 2022. http://www.theses.fr/2022UNIP7327.
Texto completoThe objective of this CIFRE thesis is to develop an image search engine, based on computer vision, to assist customs officers. Indeed, we observe, paradoxically, an increase in security threats (terrorism, trafficking, etc.) coupled with a decrease in the number of customs officers. The images of cargoes acquired by X-ray scanners already allow the inspection of a load without requiring the opening and complete search of a controlled load. By automatically proposing similar images, such a search engine would help the customs officer in his decision making when faced with infrequent or suspicious visual signatures of products. Thanks to the development of modern artificial intelligence (AI) techniques, our era is undergoing great changes: AI is transforming all sectors of the economy. Some see this advent of "robotization" as the dehumanization of the workforce, or even its replacement. However, reducing the use of AI to the simple search for productivity gains would be reductive. In reality, AI could allow to increase the work capacity of humans and not to compete with them in order to replace them. It is in this context, the birth of Augmented Intelligence, that this thesis takes place. This manuscript devoted to the question of visual similarity is divided into two parts. Two practical cases where the collaboration between Man and AI is beneficial are proposed. In the first part, the problem of learning representations for the retrieval of similar images is still under investigation. After implementing a first system similar to those proposed by the state of the art, one of the main limitations is pointed out: the semantic bias. Indeed, the main contemporary methods use image datasets coupled with semantic labels only. The literature considers that two images are similar if they share the same label. This vision of the notion of similarity, however fundamental in AI, is reductive. It will therefore be questioned in the light of work in cognitive psychology in order to propose an improvement: the taking into account of visual similarity. This new definition allows a better synergy between the customs officer and the machine. This work is the subject of scientific publications and a patent. In the second part, after having identified the key components allowing to improve the performances of thepreviously proposed system, an approach mixing empirical and theoretical research is proposed. This secondcase, augmented intelligence, is inspired by recent developments in mathematics and physics. First applied tothe understanding of an important hyperparameter (temperature), then to a larger task (classification), theproposed method provides an intuition on the importance and role of factors correlated to the studied variable(e.g. hyperparameter, score, etc.). The processing chain thus set up has demonstrated its efficiency byproviding a highly explainable solution in line with decades of research in machine learning. These findings willallow the improvement of previously developed solutions
Tirilly, Pierre. "Traitement automatique des langues pour l'indexation d'images". Phd thesis, Université Rennes 1, 2010. http://tel.archives-ouvertes.fr/tel-00516422.
Texto completoTirilly, Pierre. "Traitement automatique des langues pour l'indexation d'images". Phd thesis, Rennes 1, 2010. http://www.theses.fr/2010REN1S045.
Texto completoIn this thesis, we propose to integrate natural language processing (NLP) techniques in image indexing systems. We first address the issue of describing the visual content of images. We rely on the visual word-based image description, which raises problems that are well known in the text indexing field. First, we study various NLP methods (weighting schemes and stop-lists) to automatically determine which visual words are relevant to describe the images. Then we use language models to take account of some geometrical relations between the visual words. We also address the issue of describing the semantic content of images: we propose an image annotation scheme that relies on extracting relevant named entities from texts coming with the images to annotate
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.
Texto completoThis 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
Awad, Dounia. "Vers un système perceptuel de reconnaissance d'objets". Thesis, La Rochelle, 2014. http://www.theses.fr/2014LAROS017/document.
Texto completoThe main objective of this thesis is to propose a pipeline for an object recognition algorithm, near to human perception, and at the same time, address the problems of Content Based image retrieval (CBIR) algorithm complexity : query run time and memory allocation. In this context, we propose a filter based on visual attention system to select salient points according to human interests from the interest points extracted by a traditionnal interest points detectors. The test of our approach, using Perreira Da Silva’s system as filter, on VOC 2005 databases, demonstrated that we can maintain approximately the same performance of a object recognition system by selecting only 40% of interest points (extracted by Harris-Laplace and Laplacian), while having an important gain in complexity (40% gain in query-run time and 60% in complexity). Furthermore, we address the problem of high dimensionality of descriptor in object recognition system. We proposed a new hybrid texture descriptor, representing the spatial frequency of some perceptual features extracted by a visual attention system. This descriptor has the advantage of being lower dimension vs. traditional descriptors. Evaluating our descriptor with an object recognition system (interest points detectors are Harris-Laplace & Laplacian) on VOC 2007 databases showed a slightly decrease in the performance (with 5% loss in Average Precision) compared to the original system, based on SIFT descriptor (with 50% complexity gain). In addition, we evaluated our descriptor using a visual attention system as interest point detector, on VOC 2005 databases. The experiment showed a slightly decrease in performance (with 3% loss in performance), meanwhile we reduced drastically the complexity of the system (with 50% gain in run-query time and 70% in complexity)
Nguyen, Thanh-Khoa. "Image segmentation and extraction based on pixel communities". Thesis, La Rochelle, 2019. http://www.theses.fr/2019LAROS035.
Texto completoImage segmentation has become an indispensable task that is widely employed in several image processing applications including object detection, object tracking, automatic driver assistance, and traffic control systems, etc. The literature abounds with algorithms for achieving image segmentation tasks. These methods can be divided into some main groups according to the underlying approaches, such as Region-based image segmentation, Feature-based clustering, Graph-based approaches and Artificial Neural Network-based image segmentation. Recently, complex networks have mushroomed both theories and applications as a trend of developments. Hence, image segmentation techniques based on community detection algorithms have been proposed and have become an interesting discipline in the literature. In this thesis, we propose a novel framework for community detection based image segmentation. The idea that brings social networks analysis domain into image segmentation quite satisfies with most authors and harmony in those researches. However, how community detection algorithms can be applied in image segmentation efficiently is a topic that has challenged researchers for decades. The contribution of this thesis is an effort to construct best complex networks for applying community detection and proposal novel agglomerate methods in order to aggregate homogeneous regions producing good image segmentation results. Besides, we also propose a content based image retrieval system using the same features than the ones obtained by the image segmentation processes. The proposed image search engine for real images can implement to search the closest similarity images with query image. This content based image retrieval relies on the incorporation of our extracted features into Bag-of-Visual-Words model. This is one of representative applications denoted that image segmentation benefits several image processing and computer visions applications. Our methods have been tested on several data sets and evaluated by many well-known segmentation evaluation metrics. The proposed methods produce efficient image segmentation results compared to the state of the art
Trad, Riadh. "Découverte d'évènements par contenu visuel dans les médias sociaux". Electronic Thesis or Diss., Paris, ENST, 2013. http://www.theses.fr/2013ENST0030.
Texto completoThe ease of publishing content on social media sites brings to the Web an ever increasing amount of user generated content captured during, and associated with, real life events. Social media documents shared by users often reflect their personal experience of the event. Hence, an event can be seen as a set of personal and local views, recorded by different users. These event records are likely to exhibit similar facets of the event but also specific aspects. By linking different records of the same event occurrence we can enable rich search and browsing of social media events content. Specifically, linking all the occurrences of the same event would provide a general overview of the event. In this dissertation we present a content-based approach for leveraging the wealth of social media documents available on the Web for event identification and characterization. To match event occurrences in social media, we develop a new visual-based method for retrieving events in huge photocollections, typically in the context of User Generated Content. The main contributions of the thesis are the following : (1) a new visual-based method for retrieving events in photo collections, (2) a scalable and distributed framework for Nearest Neighbors Graph construction for high dimensional data, (3) a collaborative content-based filtering technique for selecting relevant social media documents for a given event
Trad, Riadh. "Découverte d'évènements par contenu visuel dans les médias sociaux". Thesis, Paris, ENST, 2013. http://www.theses.fr/2013ENST0030/document.
Texto completoThe ease of publishing content on social media sites brings to the Web an ever increasing amount of user generated content captured during, and associated with, real life events. Social media documents shared by users often reflect their personal experience of the event. Hence, an event can be seen as a set of personal and local views, recorded by different users. These event records are likely to exhibit similar facets of the event but also specific aspects. By linking different records of the same event occurrence we can enable rich search and browsing of social media events content. Specifically, linking all the occurrences of the same event would provide a general overview of the event. In this dissertation we present a content-based approach for leveraging the wealth of social media documents available on the Web for event identification and characterization. To match event occurrences in social media, we develop a new visual-based method for retrieving events in huge photocollections, typically in the context of User Generated Content. The main contributions of the thesis are the following : (1) a new visual-based method for retrieving events in photo collections, (2) a scalable and distributed framework for Nearest Neighbors Graph construction for high dimensional data, (3) a collaborative content-based filtering technique for selecting relevant social media documents for a given event
Omhover, Jean-François. "Recherche d'images par similarité de contenus régionaux". Paris 6, 2004. http://www.theses.fr/2004PA066254.
Texto completoHouissa, Mohamed Hichem. "Recherche par thésaurus visuel et composition spatiale dans les bases d'images". Paris 11, 2007. http://www.theses.fr/2007PA11A001.
Texto completoThe choice of the starting example is an important issue for content-based image retrieval approaches. Usual systems suggest to the user to look for images similar to the one he selected either among the database itself or from an external image collection; the results are retrieved according to specific metrics suitable with extracted descriptors. In this work, we investigated the case of a missing or at least inappropriate starting example and hence the need of mental image composition in order to initiate the search process. To do so, the paradigm of Visual Thesaurus stands for a visual summary of all regions of the database, these segmented regions are clustered into coherent categories from which we select the representatives to compose the initial "page zero". Our interest was oriented toward the construction of a reliable visual thesaurus that meets the requirements of coarse segmentation and wide variability in region's photometric and structural complexity. Global attributes are suitable to likely homogenous regions whereas fine local descriptors through Harris points of interest infer robustness and visual coherence to the categorization step. The clustering requires, on the one hand, fuzzy agglomerative algorithms but also, in case of textured patterns, relational dual formulation depending mainly on the dimension of the description space. The objective of our work is to provide an alternative to starting example by composing the mental query through the arrangement of the visual patches selected from the Visual Thesaurus. Pairs of regions are described by a weighted angular spatial histogram to determine the orientation between an argument region and a referent one. Accordingly, both logical and spatial compositions are involved; returned results rely on inverted files indexation and histogram intersection metrics respectively
Hamroun, Mohamed. "Indexation et recherche par contenu visuel, sémantique et multi-niveaux des documents multimédia". Thesis, Bordeaux, 2019. http://www.theses.fr/2019BORD0372.
Texto completoDue 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
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.
Texto completoLe, Cacheux Yannick. "Toward more practical zero-shot learning". Electronic Thesis or Diss., Paris, CNAM, 2020. http://www.theses.fr/2020CNAM1282.
Texto completoThis thesis focuses on zero-shot visual recognition, which aims to recognize images from unseen categories, i.e. categories not seen by the model during training. After categorizing existing methods into three main families, we argue that ranking methods habitually make several detrimental implicit assumptions. We propose to adapt the usual formulation of the hinge rank loss so that such methods may take inter and intra-class relations into account. We also propose a simple process to address the gap between accuracies on seen and unseen classes, from which these methods frequently suffer in a generalized zero-shot learning setting. In our experimental evaluation, the combination of these contributions enables our proposed model to equal or surpass the performance of generative methods, while being arguably less restrictive. In a second part, we focus on the semantic representations used in a large-scale zero-shot learning setting. In this setting, semantic information customarily comes from word embeddings of the class names. We argue that usual embeddings suffer from a lack of visual content in training corpora. We thus propose new visually oriented text corpora as well as a method to adapt word embedding models to these corpora. We further propose to complete unsupervised representations with short descriptions in natural language, whose generation requires minimal effort when compared to extensive attributes
Hafiane, Adel. "CARACTERISATION DE TEXTURES ET SEGMENTATION POUR LA RECHERCHE D'IMAGES PAR LE CONTENU". Phd thesis, Université Paris Sud - Paris XI, 2005. http://tel.archives-ouvertes.fr/tel-00097977.
Texto completoHlaoui, Adel. "Contribution en appariement de graphes pour la recherche d'images par le contenu". Thèse, Université de Sherbrooke, 2004. http://savoirs.usherbrooke.ca/handle/11143/5050.
Texto completoKsantini, Riadh. "Recherche d'images par le contenu, analyse multirésolution et modèles de régression logistique". Thèse, Université de Sherbrooke, 2007. http://savoirs.usherbrooke.ca/handle/11143/5088.
Texto completoFournier, 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.
Texto completoThis 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
Manjarrez, Sanchez Jorge Roberto. "Recherche par le contenu efficiente dans les bases de données parallèles d'images". Nantes, 2009. http://www.theses.fr/2009NANT2089.
Texto completoIn this thesis, we address the performance problem when searching in large databases of images. The processing of similarity queries is a computational challenge because of the dimensionality of the abstract representation for the images and size of the databases. We present two data organization methods that account for performance improvement. The first one is based on the clustering of the database in centralized settings. We derive an optimal range of values for the number of clusters to obtain from a database, which in conjunction with a searching algorithm allows to efficiently process nearest neighbor queries. However as the dimensionality and size of the database increase, a single computer is overwhelmed. The second method is based on data partitioning over a shared nothing machine. Based on the results of the first method, this method maximizes parallelism. We also derive the optimal number of processing nodes to maximize resource utilization. We performed extensive experiments with synthetic and real databases. They validate the proposals and show that the performance level is superior to existing approaches which beyond a certain dimensionality or database size become inefficient
Hafiane, Adel. "Caractérisation de textures et segmentation pour la recherche d'images par le contenu". Paris 11, 2005. http://www.theses.fr/2005PA112339.
Texto completoThis thesis describes the design and realization of a complete processing chain for content based image retrieval (CBIR). The study allows to define some limited semantics with respect to the user's satisfaction from the system response. The image is decomposed on visual entities to obtain interactions between them, allowing to reach higher levels of abstraction. We have addressed three points in the chain : reliable region-detection, region characterization and then similarity measure. We have modified a Fuzzy C-means by incorporating the spatial and multiresolution information into the objective function. Therefore, the classification of a given point is forced to follow both neighbors and ancestors in a pyramidal representation. Two methods are proposed which exploit Peano scans to coding region features. The first one is based on a grammatical representation of the pixels neighborhood called motif. The second method modifies the spectrum before to apply Gabor filters. The image signature consists of a list of visual entities containing features. The similarity measure between two images turns into a graph matching problem. We have elaborated a technique that allows a bidirectional matching from query to target and vice versa. A high priority is assigned to those elements which prefer mutually. Each part of the system is evaluated and tested independently then incorporated into the CBIR application. The evaluation of CBIR in terms of "recall-precision" shows that the proposed methods perform better than classical ones, such as grey level co-occurrence matrix and Gabor filters. To open on further extensions and suggest the generality of out method, the conclusion deals with extending it to the situation assessment in car driving, with limited tuning of parameters
Zhou, Zhyiong. "Recherche d'images par le contenu application à la proposition de mots clés". Thesis, Poitiers, 2018. http://www.theses.fr/2018POIT2254.
Texto completoThe search for information in masses of multimedia data and the indexing of these large databases by the content are very current problems. They are part of a type of data management called Digital Asset Management (or DAM) ; The DAM uses image segmentation and data classification techniques.Our main contributions in this thesis can be summarized in three points : - Analysis of the possible uses of different methods of extraction of local characteristics using the VLAD technique.- Proposed a new method for extracting dominant color information in an image.- Comparison of Support Vector Machines (SVM) to different classifiers for the proposed indexing keywords. These contributions have been tested and validated on summary data and on actual data. Our methods were then widely used in the DAM ePhoto system developed by the company EINDEN, which financed the CIFRE thesis in which this work was carried out. The results are encouraging and open new perspectives for research
Souvannavong, Fabrice. "Indexation et recherche de plans videos par le contenu sémantique". Paris, ENST, 2005. http://www.theses.fr/2005ENST0018.
Texto completoIn this thesis, we address the fussy problem of video content indexing and retrieval and in particular automatic semantic video content indexing. Indexing is the operation that consists in extracting a numerical or textual signature that describes the content in an accurate and concise manner. The objective is to allow an efficient search in a database. The automatic aspect of the indexing is important since we can imagine the difficulty to annotate video shots in huge databases. Until now, systems were concentrated on the description and indexing of the visual content. The search was mainly led on colors and textures of video shots. The new challenge is now to automatically add to these signatures a semantic description of the content. First, a range of indexing techniques is presented. Second, we introduce a method to compute an accurate and compact signature from key-frames regions. This method is an adaptation of the latent semantic indexing method originally used to index text documents. Third, we address the difficult task of semantic content retrieval. Experiments are led in the framework of TRECVID. It allows having a huge amount of videos and their labels. Fourth, we pursue on the semantic classification task through the study of fusion mechanisms. Finally, this thesis concludes on the introduction of a new active learning approach to limit the annotation effort
Letessier, Pierre. "Découverte et exploitation d'objets visuels fréquents dans des collections multimédia". Thesis, Paris, ENST, 2013. http://www.theses.fr/2013ENST0014/document.
Texto completoThe main goal of this thesis is to discover frequent visual objects in large multimedia collections. As in many areas (finance, genetics, . . .), it consists in extracting a knowledge, using the occurence frequency of an object in a collection as a relevance criterion. A first contribution is to provide a formalism to the problems of mining and discovery of frequent visual objects. The second contribution is a generic method to solve these two problems, based on an iterative sampling process, and on an efficient and scalable rigid objects matching. The third contribution of this work focuses on building a likelihood function close to the perfect distribution. Experiments show that contrary to state-of-the-art methods, our approach allows to discover efficiently very small objects in several millions images. Finally, several applications are presented, including trademark logos discovery, transmedia events detection or visual-based query suggestion
Ordon︢ez, Varela John Richard. "Indexation et recherche d'images par le contenu, utilisant des informations de compression d'images : application aux images médicales". Rennes 1, 2004. http://www.theses.fr/2004REN10009.
Texto completoDa, Rugna Jérôme. "De l'usage des méthodes bas niveau pour la recherche d'images par le contenu". Saint-Etienne, 2004. http://www.theses.fr/2004STET4015.
Texto completoThe matter of this work is content based image retrievaland more precisely the contribution of the low level methods. After having discussed the various existing approaches, we recall the semantic gap between the user expectations and what really the systems of research propose. Most of these approaches rely on a preliminary step of segmentation whose validity and robustness must be studied. Then we propose a protocol of evaluation and a practical example of benchmarks. The originality consists in not comparing a segmentation with a theoretical reference but judging its stability objectively. The third part of this document introduces three specific contributions likely to improve the chain of research. Initially, a detector of blur allows to extract a meta-data carried by the image: the unblur regions, a priori of focusing. Secondly, we expose a descriptor based on the extraction of emergent areas using only the color criteria. This process, combined with adapted distances, may allow for example a color pre-filtering before the step of similarity research. Finally, we briefly introduce an algebra of histograms able as well as possible to exploit the information contained in this type of descriptors, via a specific query language
Abbadeni, Noureddine. "Recherche d'images basée sur le contenu visuel : représentations multiples, similarité et fusion de résultats : cas des images de texture". Thèse, Université de Sherbrooke, 2005. http://savoirs.usherbrooke.ca/handle/11143/5045.
Texto completoQiao, Yongliang. "Place recognition based visual localization in changing environments". Thesis, Bourgogne Franche-Comté, 2017. http://www.theses.fr/2017UBFCA004/document.
Texto completoIn many applications, it is crucial that a robot or vehicle localizes itself within the world especially for autonomous navigation and driving. The goal of this thesis is to improve place recognition performance for visual localization in changing environment. The approach is as follows: in off-line phase, geo-referenced images of each location are acquired, features are extracted and saved. While in the on-line phase, the vehicle localizes itself by identifying a previously-visited location through image or sequence retrieving. However, visual localization is challenging due to drastic appearance and illumination changes caused by weather conditions or seasonal changing. This thesis addresses the challenge of improving place recognition techniques through strengthen the ability of place describing and recognizing. Several approaches are proposed in this thesis:1) Multi-feature combination of CSLBP (extracted from gray-scale image and disparity map) and HOG features is used for visual localization. By taking the advantages of depth, texture and shape information, visual recognition performance can be improved. In addition, local sensitive hashing method (LSH) is used to speed up the process of place recognition;2) Visual localization across seasons is proposed based on sequence matching and feature combination of GIST and CSLBP. Matching places by considering sequences and feature combination denotes high robustness to extreme perceptual changes;3) All-environment visual localization is proposed based on automatic learned Convolutional Network (ConvNet) features and localized sequence matching. To speed up the computational efficiency, LSH is taken to achieve real-time visual localization with minimal accuracy degradation
Daoudi, Imane. "Recherche par similarité dans les grandes bases de données multimédia : application à la recherche par le contenu dans les bases d'images". Lyon, INSA, 2009. http://theses.insa-lyon.fr/publication/2009ISAL0057/these.pdf.
Texto completo[The emergence of digital multimedia data is increasing. Access, sharing and retrieval of these data have become the real needs. This requires the use of powerful tools and search engine for fast and efficient access to data. The spectacular growth of technologies and numeric requires the use of powerful tools and search engine for fast and efficient access to data. My thesis work is in the field of multimedia data especially images. The main objectives is to develop a fast and efficient indexing and searching method of the k nearest neighbour which is adapted for applications in Content-based image retrieval (CBIR) and for properties of image descriptors (high volume, large dimension, etc. ). The main idea is on one hand, to provide answers to the problems of scalability and the curse of dimensionality and the other to deal with similarity problems that arise in indexing and CBIR. We propose in this thesis two different approaches. The first uses a multidimensional indexing structure based on approximation approach or filtering, which is an improvement in the RA-Blocks method. The proposed method is based on the proposal of an algorithm of subdividing the data space which improves the storage capacity of the index and the CPU times. In a second approach, we propose a multidimensional indexing method suitable for heterogeneous data (colour, texture, shape). The second proposed method combines a non linear dimensionality reduction technique with a multidimensional indexing approach based on approximation. This combination allows one hand to deal with the curse of dimensionality scalability problems and also to exploit the properties of the non-linear space to find suitable similarity measurement for the nature of manipulated data. ]
Hamri, Touati. "Approche probabiliste hybride pour la recherche d'images par le contenu avec pondération des caractéristiques". Mémoire, Université de Sherbrooke, 2007. http://savoirs.usherbrooke.ca/handle/11143/4784.
Texto completoRiadh, Ksantini. "Recherche d'images par le contenu, analyse multirésolution et modèles de régression logistiqueh[ressource électronique] /". [S.l. : s.n.], 2007.
Buscar texto completoTaïleb, Mounira. "NOHIS-tree nouvelle méthode de recherche de plus proches voisins : application à la recherche d'images par le contenu". Paris 11, 2008. http://www.theses.fr/2008PA112164.
Texto completoThe increasing of image databases requires the use of a content-based image retrieval system (CBIR). A such system consist first to describe automatically the images, visual properties of each image are represented as multidimensional vectors called descriptors. Next, finding similar images to the query image is achieved by searching for the nearest neighbors of each descriptor of the query image. In this thesis, we propose a new method for indexing multidimensional bases with the search algorithm of nearest neighbors adapted. The originality of our multidimensional index is the disposition of the bounding forms avoiding overlapping. Indeed, the overlapping is one of the main drawbacks that slow the search of nearest neighbors search. Our index with its search algorithm speeds the nearest neighbors search while doing an exact search. Our method has been integrated and tested within a real content-based image system. The results of tests carried out show the robustness of our method in terms of accuracy and speed in search time
Jai, Andaloussi Said. "Indexation de l'information médicale. Application à la recherche d'images et de vidéos par le contenu". Télécom Bretagne, 2010. http://www.theses.fr/2010TELB0150.
Texto completoThis PhD thesis addresses the use of multimedia medical databases for diagnostic decision and therapeutic follow-up. Our goal is to develop methods and a system to select in multimedia databases documents similar to a query document. These documents consist of text information, numeric images and sometimes videos. In the proposed diagnosis aid system, the database is queried with the patient file, or a part of it, as input. Our work therefore involves implementing methods related to Case-Based Reasoning (CBR), datamining, Content Based Image Retrieval (CBIR) and Content Based Video Retrieval (CBVR). These methods are evaluated on three multimodal medical databases. The first database consists of retinal images collected by the LaTIM laboratory for aided diabetic retinopathy follow-up. The second database is a public mammography database (Digital Database for Screening Mammography – DDSM –) collected by the University of South Florida. The third database consists of gastroenterology videos also collected by the LaTIM laboratory. This database is used to discover whether methods developed for fixed image retrieval can also be used for color video retrieval. The first part of this work focuses on the characterization of each image in the patient file. We continued the work started in our laboratory to characterize images globally in the compressed domain (vector quantization, DCT-JPEG, wavelets, adapted wavelets) for image retrieval. Compared to other compression methods, the wavelet decomposition led to a great improvement in terms of retrieval performance. However, the wavelet decomposition requires the specification of a kernel or basis function. To overcome this problem, we proposed an original image characterization method based on the BEMD (Bidimensionnal Empirical Mode Decomposition). It allows decomposing an image into several BIMFs (Bidimensionnal Intrinsic Mode Functions) that provide access to frequency information of the image content. An originality of the method comes from the self-adaptivity of BEMD: it does not require the specification of a basic function. Once images are characterized, a similarity search is performed by computing the distance between the signature of the query image and the signature of each image in the database, given a metric. This process leads to the selection of similar images, without semantic meaning. An optimization process, based on genetic algorithms, is used to adapt the distance metric and thus improve retrieval performance. Then, the problem of content based video retrieval is addressed. A method to generate video signatures is presented. This method relies on key video frames extracted by movement analysis. The distance between video signatures is computed using a Principal Component Analysis (PCA) based technique. Finally, the proposed methods are integrated into the framework of patient file retrieval (each patient file consisting of several images and textual information). Three methods developed during a PhD thesis recently defended in our laboratory are used for patient file retrieval: the first approach is based on decision trees and their extensions, the second on Bayesian networks and the third on the Dezert-Smarandache theory (DSmT).
Najjar, Micheline. "Modèles de mélange pour la recherche d'images par le contenu : Applications aux pathologies ostéo-articulaires". Compiègne, 2004. http://www.theses.fr/2004COMP1507.
Texto completoZhao, Shuji. "Catégorisation par le contenu sémantique d'objets vidéo : recherche et reconnaissance d'acteurs dans les films". Thesis, Cergy-Pontoise, 2011. http://www.theses.fr/2011CERG0511/document.
Texto completoIn this thesis, we propose a new video object retrieval and recognition system based on visual content.From video sequences, we detect, then extract video objects such as face and car, and define the continuous content made of regions containing this object in successive frames. From this volume, called Track, we extract spatio-temporally consistent visual features to define the video object representation: Spatio-Temporal Tube.To evaluate the similarity between complex tube objects, we design a Spatio-Temporal Tube Kernel (STTK) function. Based on this kernel similarity we present both supervised and active learning strategies embedded in Support Vector Machine framework. Additionally, we propose a multi-class classification framework dealing with highly unbalanced datasets.Our approach is successfully evaluated on real movie databases. Our machine learning approach outperforms the state of the art methods for multi-class actor recognition. Our method is also evaluated for actor retrieval task and on a car database showing hence promising results for car identification task and the potential of extension to any category of video objects
Berrani, Sid-Ahmed. "Recherche approximative de plus proches voisins avec contrôle probabiliste de la précision ; application à la recherche d'images par le contenu". Phd thesis, Université Rennes 1, 2004. http://tel.archives-ouvertes.fr/tel-00532854.
Texto completoKrapac, Josip. "Représentations d'images pour la recherche et la classification d'images". Phd thesis, Université de Caen, 2011. http://tel.archives-ouvertes.fr/tel-00650998.
Texto completoVieux, Rémi. "Extraction de Descripteurs Pertinents et Classification pour le Problème de Recherche des Images par le Contenu". Thesis, Bordeaux 1, 2011. http://www.theses.fr/2011BOR14244/document.
Texto completoThe explosive development of affordable, high quality image acquisition deviceshas made available a tremendous amount of digital content. Large industrial companies arein need of efficient methods to exploit this content and transform it into valuable knowledge.This PhD has been accomplished in the context of the X-MEDIA project, a large Europeanproject with two major industrial partners, FIAT for the automotive industry andRolls-Royce plc. for the aircraft industry. The project has been the trigger for research linkedwith strong industrial requirements. Although those user requirements can be very specific,they covered more generic research topics. Hence, we bring several contributions in thegeneral context of Content-Based Image Retrieval (CBIR), Indexing and Classification.In the first part of the manuscript we propose contributions based on the extraction ofglobal image descriptors. We rely on well known descriptors from the literature to proposemodels for the indexing of image databases, and the approximation of a user defined categorisation.Additionally, we propose a new descriptor for a CBIR system which has toprocess a very specific image modality, for which traditional descriptors are irrelevant. Inthe second part of the manuscript, we focus on the task of image classification. Industrialrequirements on this topic go beyond the task of global image classification. We developedtwo methods to localize and classify the local content of images, i.e. image regions, usingsupervised machine learning algorithms (Support Vector Machines). In the last part of themanuscript, we propose a model for Content-Based Image Retrieval based on the constructionof a visual dictionary of image regions. We extensively experiment the model in orderto identify the most influential parameters in the retrieval efficiency
Bouguila, Nizar. "Les mixtures de Dirichlet et leurs apports pour la classification et la recherche d'images par le contenu". Mémoire, [S.l. : s.n.], 2002. http://savoirs.usherbrooke.ca/handle/11143/4565.
Texto completoDa, Rugna Jérôme. "De l'usage des méthodes bas niveau pour la recherche d'image par le contenu". Phd thesis, Université Jean Monnet - Saint-Etienne, 2004. http://tel.archives-ouvertes.fr/tel-00070811.
Texto completol'apport des méthodes bas niveau.
Après avoir discuté des différentes approches existantes, nous rappelons le fossé sémantique
entre les attentes de l'utilisateur et ce que proposent réellement les systèmes de recherche. La
plupart de ceux-ci reposent sur une étape préalable de segmentation dont la validité et la robustesse
se doivent d'être étudiées. Nous proposons alors un protocole d'évaluation objective et un
exemple concret de mise en oeuvre. L'originalité consiste à ne pas comparer une segmentation à
une référence théorique mais à juger objectivement sa stabilité.
La troisième partie de ce document introduit trois contributions ponctuelles susceptibles
d'améliorer la chaîne de recherche. Dans un premier temps, un détecteur de flou permet d'extraire
une méta-information portée par l'image, les zones nettes a priori de focalisation. Ensuite
nous exposons un descripteur basé sur l'extraction de régions émergentes sur le seul critère couleur.
Cette extraction, conjuguée avec des distances adaptées, peut permettre par exemple un
pré-filtrage couleur en amont de la phase de recherche de similarité proprement dite. Finalement,
nous introduisons brièvement une algèbre d'histogrammes pour exploiter au mieux l'information
contenue dans ce type de descripteurs, via un langage de requêtes spécifique.
Letessier, Pierre. "Découverte et exploitation d'objets visuels fréquents dans des collections multimédia". Electronic Thesis or Diss., Paris, ENST, 2013. http://www.theses.fr/2013ENST0014.
Texto completoThe main goal of this thesis is to discover frequent visual objects in large multimedia collections. As in many areas (finance, genetics, . . .), it consists in extracting a knowledge, using the occurence frequency of an object in a collection as a relevance criterion. A first contribution is to provide a formalism to the problems of mining and discovery of frequent visual objects. The second contribution is a generic method to solve these two problems, based on an iterative sampling process, and on an efficient and scalable rigid objects matching. The third contribution of this work focuses on building a likelihood function close to the perfect distribution. Experiments show that contrary to state-of-the-art methods, our approach allows to discover efficiently very small objects in several millions images. Finally, several applications are presented, including trademark logos discovery, transmedia events detection or visual-based query suggestion
Dorval, Thierry. "Approches saillantes et psycho-visuelles pour l'indexation d'images couleurs". Paris 6, 2004. http://www.theses.fr/2004PA066096.
Texto completo