Thèses sur le sujet « Hierarchical representations of images »
Pour voir les autres types de publications sur ce sujet consultez le lien suivant : Hierarchical representations of images.
Créez une référence correcte selon les styles APA, MLA, Chicago, Harvard et plusieurs autres
Consultez les 50 meilleures thèses pour votre recherche sur le sujet « Hierarchical representations of images ».
À côté de chaque source dans la liste de références il y a un bouton « Ajouter à la bibliographie ». Cliquez sur ce bouton, et nous générerons automatiquement la référence bibliographique pour la source choisie selon votre style de citation préféré : APA, MLA, Harvard, Vancouver, Chicago, etc.
Vous pouvez aussi télécharger le texte intégral de la publication scolaire au format pdf et consulter son résumé en ligne lorsque ces informations sont inclues dans les métadonnées.
Parcourez les thèses sur diverses disciplines et organisez correctement votre bibliographie.
1
Fehri, Amin. « Image Characterization by Morphological Hierarchical Representations ». Thesis, Paris Sciences et Lettres (ComUE), 2018. http://www.theses.fr/2018PSLEM063/document.
Texte intégralRésumé :
Cette thèse porte sur l'extraction de descripteurs hiérarchiques et multi-échelles d'images, en vue de leur interprétation, caractérisation et segmentation. Elle se décompose en deux parties.La première partie expose des éléments théoriques et méthodologiques sur l'obtention de classifications hiérarchiques des nœuds d'un graphe valué aux arêtes. Ces méthodes sont ensuite appliquées à des graphes représentant des images pour obtenir différentes méthodes de segmentation hiérarchique d'images. De plus, nous introduisons différentes façons de combiner des segmentations hiérarchiques. Nous proposons enfin une méthodologie pour structurer et étudier l'espace des hiérarchies que nous avons construites en utilisant la distance de Gromov-Hausdorff entre elles.La seconde partie explore plusieurs applications de ces descriptions hiérarchiques d'images. Nous exposons une méthode pour apprendre à extraire de ces hiérarchies une bonne segmentation de façon automatique, étant donnés un type d'images et un score de bonne segmentation. Nous proposons également des descripteurs d'images obtenus par mesure des distances inter-hiérarchies, et exposons leur efficacité sur des données réelles et simulées. Enfin, nous étendons les potentielles applications de ces hiérarchies en introduisant une technique permettant de prendre en compte toute information spatiale a priori durant leur constructionThis thesis deals with the extraction of hierarchical and multiscale descriptors on images, in order to interpret, characterize and segment them. It breaks down into two parts.The first part outlines a theoretical and methodological approach for obtaining hierarchical clusterings of the nodes of an edge-weighted graph. In addition, we introduce different approaches to combine hierarchical segmentations. These methods are then applied to graphs representing images and derive different hierarchical segmentation techniques. Finally, we propose a methodology for structuring and studying the space of hierarchies by using the Gromov-Hausdorff distance as a metric.The second part explores several applications of these hierarchical descriptions for images. We expose a method to learn how to automatically extract a segmentation of an image, given a type of images and a score of evaluation for a segmentation. We also propose image descriptors obtained by measuring inter-hierarchical distances, and expose their efficiency on real and simulated data. Finally, we extend the potential applications of these hierarchies by introducing a technique to take into account any spatial prior information during their construction
2
Cui, Yanwei. « Kernel-based learning on hierarchical image representations : applications to remote sensing data classification ». Thesis, Lorient, 2017. http://www.theses.fr/2017LORIS448/document.
Texte intégralRésumé :
La représentation d’image sous une forme hiérarchique a été largement utilisée dans un contexte de classification. Une telle représentation est capable de modéliser le contenu d’une image à travers une structure arborescente. Dans cette thèse, nous étudions les méthodes à noyaux qui permettent de prendre en entrée des données sous une forme structurée et de tenir compte des informations topologiques présentes dans chaque structure en concevant des noyaux structurés. Nous présentons un noyau structuré dédié aux structures telles que des arbres non ordonnés et des chemins (séquences de noeuds) équipés de caractéristiques numériques. Le noyau proposé, appelé Bag of Subpaths Kernel (BoSK), est formé en sommant les noyaux calculés sur les sous-chemins (un sac de tous les chemins et des noeuds simples) entre deux sacs. Le calcul direct de BoSK amène à une complexité quadratique par rapport à la taille de la structure (nombre de noeuds) et la quantité de données (taille de l’ensemble d’apprentissage). Nous proposons également une version rapide de notre algorithme, appelé Scalable BoSK (SBoSK), qui s’appuie sur la technique des Random Fourier Features pour projeter les données structurées dans un espace euclidien, où le produit scalaire du vecteur transformé est une approximation de BoSK. Cet algorithme bénéficie d’une complexité non plus linéaire mais quadratique par rapport aux tailles de la structure et de l’ensemble d’apprentissage, rendant ainsi le noyau adapté aux situations d’apprentissage à grande échelle. Grâce à (S)BoSK, nous sommes en mesure d’effectuer un apprentissage à partir d’informations présentes à plusieurs échelles dans les représentations hiérarchiques d’image. (S)BoSK fonctionne sur des chemins, permettant ainsi de tenir compte du contexte d’un pixel (feuille de la représentation hiérarchique) par l’intermédiaire de ses régions ancêtres à plusieurs échelles. Un tel modèle est utilisé dans la classification des images au niveau pixel. (S)BoSK fonctionne également sur les arbres, ce qui le rend capable de modéliser la composition d’un objet (racine de la représentation hiérarchique) et les relations topologiques entre ses sous-parties. Cette stratégie permet la classification des tuiles ou parties d’image. En poussant plus loin l’utilisation de (S)BoSK, nous introduisons une nouvelle approche de classification multi-source qui effectue la classification directement à partir d’une représentation hiérarchique construite à partir de deux images de la même scène prises à différentes résolutions, éventuellement selon différentes modalités. Les évaluations sur plusieurs jeux de données de télédétection disponibles dans la communauté illustrent la supériorité de (S)BoSK par rapport à l’état de l’art en termes de précision de classification, et les expériences menées sur une tâche de classification urbaine montrent la pertinence de l’approche de classification multi-source proposéeHierarchical image representations have been widely used in the image classification context. Such representations are capable of modeling the content of an image through a tree structure. In this thesis, we investigate kernel-based strategies that make possible taking input data in a structured form and capturing the topological patterns inside each structure through designing structured kernels. We develop a structured kernel dedicated to unordered tree and path (sequence of nodes) structures equipped with numerical features, called Bag of Subpaths Kernel (BoSK). It is formed by summing up kernels computed on subpaths (a bag of all paths and single nodes) between two bags. The direct computation of BoSK yields a quadratic complexity w.r.t. both structure size (number of nodes) and amount of data (training size). We also propose a scalable version of BoSK (SBoSK for short), using Random Fourier Features technique to map the structured data in a randomized finite-dimensional Euclidean space, where inner product of the transformed feature vector approximates BoSK. It brings down the complexity from quadratic to linear w.r.t. structure size and amount of data, making the kernel compliant with the large-scale machine-learning context. Thanks to (S)BoSK, we are able to learn from cross-scale patterns in hierarchical image representations. (S)BoSK operates on paths, thus allowing modeling the context of a pixel (leaf of the hierarchical representation) through its ancestor regions at multiple scales. Such a model is used within pixel-based image classification. (S)BoSK also works on trees, making the kernel able to capture the composition of an object (top of the hierarchical representation) and the topological relationships among its subparts. This strategy allows tile/sub-image classification. Further relying on (S)BoSK, we introduce a novel multi-source classification approach that performs classification directly from a hierarchical image representation built from two images of the same scene taken at different resolutions, possibly with different modalities. Evaluations on several publicly available remote sensing datasets illustrate the superiority of (S)BoSK compared to state-of-the-art methods in terms of classification accuracy, and experiments on an urban classification task show the effectiveness of proposed multi-source classification approach
3
Lagrange, Adrien. « From representation learning to thematic classification - Application to hierarchical analysis of hyperspectral images ». Thesis, Toulouse, INPT, 2019. http://www.theses.fr/2019INPT0095.
Texte intégralRésumé :
De nombreuses approches ont été développées pour analyser la quantité croissante de donnée image disponible. Parmi ces méthodes, la classification supervisée a fait l'objet d'une attention particulière, ce qui a conduit à la mise au point de méthodes de classification efficaces. Ces méthodes visent à déduire la classe de chaque observation en se basant sur une nomenclature de classes prédéfinie et en exploitant un ensemble d'observations étiquetées par des experts. Grâce aux importants efforts de recherche de la communauté, les méthodes de classification sont devenues très précises. Néanmoins, les résultats d'une classification restent une interprétation haut-niveau de la scène observée puisque toutes les informations contenues dans une observation sont résumées en une unique classe. Contrairement aux méthodes de classification, les méthodes d'apprentissage de représentation sont fondées sur une modélisation des données et conçues spécialement pour traiter des données de grande dimension afin d'en extraire des variables latentes pertinentes. En utilisant une modélisation basée sur la physique des observations, ces méthodes permettent à l'utilisateur d'extraire des variables très riches de sens et d'obtenir une interprétation très fine de l'image considérée. L'objectif principal de cette thèse est de développer un cadre unifié pour l'apprentissage de représentation et la classification. Au vu de la complémentarité des deux méthodes, le problème est envisagé à travers une modélisation hiérarchique. L'approche par apprentissage de représentation est utilisée pour construire un modèle bas-niveau des données alors que la classification, qui peut être considérée comme une interprétation haut-niveau des données, est utilisée pour incorporer les informations supervisées. Deux paradigmes différents sont explorés pour mettre en place ce modèle hiérarchique, à savoir une modélisation bayésienne et la construction d'un problème d'optimisation. Les modèles proposés sont ensuite testés dans le contexte particulier de l'imagerie hyperspectrale où la tâche d'apprentissage de représentation est spécifiée sous la forme d'un problème de démélange spectralNumerous frameworks have been developed in order to analyze the increasing amount of available image data. Among those methods, supervised classification has received considerable attention leading to the development of state-of-the-art classification methods. These methods aim at inferring the class of each observation given a specific class nomenclature by exploiting a set of labeled observations. Thanks to extensive research efforts of the community, classification methods have become very efficient. Nevertheless, the results of a classification remains a highlevel interpretation of the scene since it only gives a single class to summarize all information in a given pixel. Contrary to classification methods, representation learning methods are model-based approaches designed especially to handle high-dimensional data and extract meaningful latent variables. By using physic-based models, these methods allow the user to extract very meaningful variables and get a very detailed interpretation of the considered image. The main objective of this thesis is to develop a unified framework for classification and representation learning. These two methods provide complementary approaches allowing to address the problem using a hierarchical modeling approach. The representation learning approach is used to build a low-level model of the data whereas classification is used to incorporate supervised information and may be seen as a high-level interpretation of the data. Two different paradigms, namely Bayesian models and optimization approaches, are explored to set up this hierarchical model. The proposed models are then tested in the specific context of hyperspectral imaging where the representation learning task is specified as a spectral unmixing problem
4
Xu, Zijian. « A hierarchical compositional model for representation and sketching of high-resolution human images ». Diss., Restricted to subscribing institutions, 2007. http://proquest.umi.com/pqdweb?did=1495960431&sid=1&Fmt=2&clientId=1564&RQT=309&VName=PQD.
Texte intégral
5
Huynh, Lê Duy. « Taking into account inclusion and adjacency information in morphological hierarchical representations, with application to the extraction of text in natural images and videos ». Electronic Thesis or Diss., Sorbonne université, 2018. http://www.theses.fr/2018SORUS341.
Texte intégralRésumé :
Les relations d'inclusion et d'adjacence des regions dans l'images comportent des informations contextuelles. Le relation d'adjacence est largement utilisé car il indique comment les régions sont organisées dans l'images. La relation d'inclusion n'est généralement pas prise en compte, bien qu'il assimile la relation d'objet-fond. Il existe plusieurs représentations morphologiques hiérarchiques: l'arbre des formes (AdF) qui représentent l'inclusion de lignes de niveaux d'image, ainsi que les hiérarchies de segmentation (i.e. la hiérarchie des quasi-zones plates) qui est utile dans l'analyse de la relation d'adjacence. Le but de ce travail est de tirer partie à la fois des relations d’inclusion et d’adjacence dans ces representations pour mener à bien des tâches de vision par ordinateur. Nous introduisons le graphe d'alignement spatial (GAS) qui est construit à partir de l'inclusion et de l'arrangement spatial des régions dans l'AdF. Dans un cas simple tel que notre l'AdF de Laplacien, le GAS est réduit à un graphe déconnecté où chaque composant connecté est un groupe sémantique d'objets. Dans d’autres cases, e.g., l'AdF classique, le GAS est plus complexe. Pour résoudre ce problème, nous proposons d'élargir notre raisonnement à la morphologie basée sur la forme. Notre extension permet de manipuler n'importe quel graphe des formes et permet n'importe stratégie de filtrage dans la cadre de opérateurs connexes. Par conséquent, le GAS pourrait être analysé par une hiérarchie des quasi-zones plates. Les résultats de notre méthode dans la reconnaissance de texte montrent l'efficacité et la performance, qui sont attrayantes notablement pour les applications mobilesThe inclusion and adjacency relationship between image regions usually carry contextual information. The later is widely used since it tells how regions are arranged in images. The former is usually not taken into account although it parallels the object-background relationship. The mathematical morphology framework provides several hierarchical image representations. They include the Tree of Shapes (ToS), which encodes the inclusion of level-line, and the hierarchies of segmentation (e.g., alpha-tree, BPT), which is useful in the analysis of the adjacency relationship. In this work, we take advantage of both inclusion and adjacency information in these representations for computer vision applications. We introduce the spatial alignment graph w.r.t inclusion that is constructed by adding a new adjacency relationship to nodes of the ToS. In a simple ToS such as our Tree of Shapes of Laplacian sign, which encodes the inclusion of Morphological Laplacian 0-crossings, the graph is reduced to a disconnected graph where each connected component is a semantic group. In other cases, e.g., classic ToS, the spatial alignment graph is more complex. To address this issue, we expand the shape-spaces morphology. Our expansion has two primary results: 1)It allows the manipulation of any graph of shapes. 2)It allows any tree filtering strategy proposed by the connected operators frameworks. With this expansion, the spatial graph could be analyzed with the help of an alpha-tree. We demonstrated the application aspect of our method in the application of text detection. The experiment results show the efficiency and effectiveness of our methods, which is appealing to mobile applications
6
Esteban, Baptiste. « A Generic, Efficient, and Interactive Approach to Image Processing with Applications in Mathematical Morphology ». Electronic Thesis or Diss., Sorbonne université, 2023. http://www.theses.fr/2023SORUS623.
Texte intégralRésumé :
Les bibliothèques de traitement d’images jouent un rôle important dans la boîte à outils du chercheur et devraient respecter trois critères : généricité, performance et interactivité. La généricité favorise la réutilisation du code et la flexibilité des algorithmes pour diverses structures de données en entrée, tandis que la performance accélère les expériences et permet l’utilisation d’algorithmes dans le cas d’applications en temps réel. De plus, l’interactivité dans la chaîne de traitement d’une image permet d’effectuer des expérimentations en échangeant des données avec cette dernière. Ce dernier critère est généralement obtenu en ajoutant du dynamisme à la bibliothèque, et plus particulièrement en interfaçant ses fonctionnalités à un langage dynamique. Les deux premiers critères peuvent être atteints avec des langages statiques tels que C++ ou Rust, qui exigent la connaissance de certaines informations au moment de la compilation pour optimiser le code machine généré en fonction des différents types de données d’entrée et de sortie d’un algorithme. Le dernier critère nécessite généralement d’attendre jusqu’à l’exécution pour obtenir des informations sur le type, et est donc réalisé au détriment de la vitesse d’exécution. Le travail présenté dans cette thèse vise à dépasser cette limitation dans le contexte d’algorithmes de traitement d’images. Pour ce faire, une méthodologie visant à développer des algorithmes génériques dont les informations sur les types d’entrée et de sortie peuvent être connues soit au moment de la compilation, soit à l’exécution, est présentée. Cette méthode est évaluée sur différents schémas algorithmiques de traitement d’images, et il est conclu que l’écart de performance entre les versions où l’information de type est connu à la compilation et à l’exécution de l’algorithme de construction pour les représentations hiérarchiques d’images est négligeable. En tant qu’application, les représentations hiérarchiques sont utilisées pour étendre l’applicabilité de l’estimation du niveau de bruit en niveaux de gris aux images en couleur afin d’améliorer leur caractère générique. Cela soulève l’importance d’étudier l’impact d’une telle altération dans les images à partir desquelles les représentations hiérarchiques sont construites pour améliorer l’efficacité de leurs applications en présence de bruit. Il est démontré que le bruit a un impact sur la structure arborescente, et cet impact est lié à certains types de fonctionnelles dans le cas où les hiérarchies sont contraintes par une énergieImage processing libraries play an important role in the researcher toolset and should respect three criteria: genericity, performance, and interactivity. In short, genericity boosts code reuse and algorithm flexibility for various data inputs, while performance speeds up experiments and supports real-time applications. Additionally, interactivity allows software evolution and maintenance without full recompilation, often through integration with dynamic languages like Python or Julia. The first two criteria are not straightforward to reach with static languages such as C++ or Rust which require knowing some information at compile time to optimize generated machine code related to the different input and output data types of an algorithm. The latest criterion usually requires waiting until runtime to obtain type information and is thus performed at the cost of runtime efficiency. The work presented in this thesis aims to go beyond this limitation in the context of image processing algorithms. To do so, a methodology to develop generic algorithms whose type information about its input and output data may be known either at compile-time or at runtime is presented. This methodology is evaluated on different image processing algorithmic schemes, and it is concluded that the performance gap between the runtime and compile-time versions of the construction algorithm for hierarchical representations of images is negligible. As an application, hierarchical representations are employed to expand the applicability of grayscale noise level estimation to color images to enhance its genericity. That raises the importance of studying the impact of such corruption in the hierarchies built on noisy images to improve their efficiency in the presence of noise. It is demonstrated that the noise has an impact on the tree structure, and this impact is related to some kinds of functional in the context of energy optimization on hierarchies
7
Drumetz, Lucas. « Endmember Variability in hyperspectral image unmixing ». Thesis, Université Grenoble Alpes (ComUE), 2016. http://www.theses.fr/2016GREAT075/document.
Texte intégralRésumé :
La finesse de la résolution spectrale des images hyperspectrales en télédétection permet une analyse précise de la scène observée, mais leur résolution spatiale est limitée, et un pixel acquis par le capteur est souvent un mélange des contributions de différents matériaux. Le démélange spectral permet d'estimer les spectres des matériaux purs (endmembers) de la scène, et leurs abondances dans chaque pixel. Les endmembers sont souvent supposés être parfaitement représentés par un seul spectre, une hypothèse fausse en pratique, chaque matériau ayant une variabilité intra-classe non négligeable. Le but de cette thèse est de développer des algorithmes prenant mieux en compte ce phénomène. Nous effectuons le démélange localement, dans des régions bien choisies de l'image où les effets de la variabilité sont moindres, en éliminant automatiquement les endmembers non pertinents grâce à de la parcimonie collaborative. Dans une autre approche, nous raffinons l'estimation des abondances en utilisant la structure de groupe d'un dictionnaire d'endmembers extrait depuis les données. Ensuite, nous proposons un modèle de mélange linéaire étendu, basé sur des considérations physiques, qui modélise la variabilité spectrale par des facteurs d'échelle, et développons des algorithmes d'optimisation pour en estimer les paramètres. Ce modèle donne des résultats facilement interprétables et de meilleures performances que d'autres approches de la littérature. Nous étudions enfin deux applications de ce modèle pour confirmer sa pertinenceThe fine spectral resolution of hyperspectral remote sensing images allows an accurate analysis of the imaged scene, but due to their limited spatial resolution, a pixel acquired by the sensor is often a mixture of the contributions of several materials. Spectral unmixing aims at estimating the spectra of the pure materials (called endmembers) in the scene, and their abundances in each pixel. The endmembers are usually assumed to be perfectly represented by a single spectrum, which is wrong in practice since each material exhibits a significant intra-class variability. This thesis aims at designing unmixing algorithms to better handle this phenomenon. First, we perform the unmixing locally in well chosen regions of the image where variability effects are less important, and automatically discard wrongly estimated local endmembers using collaborative sparsity. In another approach, we refine the abundance estimation of the materials by taking into account the group structure of an image-derived endmember dictionary. Second, we introduce an extended linear mixing model, based on physical considerations, modeling spectral variability in the form of scaling factors, and develop optimization algorithms to estimate its parameters. This model provides easily interpretable results and outperforms other state-of-the-art approaches. We finally investigate two applications of this model to confirm its relevance
8
Yeh, Hur-jye. « 3-D reconstruction and image encoding using an efficient representation of hierarchical data structure / ». The Ohio State University, 1987. http://rave.ohiolink.edu/etdc/view?acc_num=osu148732651171353.
Texte intégral
9
Keeter, Matthew (Matthew Joseph). « Hierarchical volumetric object representations for digital fabrication workflows ». Thesis, Massachusetts Institute of Technology, 2013. http://hdl.handle.net/1721.1/82426.
Texte intégralRésumé :
Thesis (S.M.)--Massachusetts Institute of Technology, School of Architecture and Planning, Program in Media Arts and Sciences, 2013.Cataloged from PDF version of thesis.
Includes bibliographical references (p. 111-114).
Modern systems for computer-aided design and manufacturing (CAD/CAM) have a history dating back to drafting boards, early computers, and machine shops with specialized technicians for each stage in a manufacturing workflow. In recent years, personal-scale digital fabrication has challenged many of these workflows' build-in assumptions. A single individual may control the entire workflow, from design to manufacture; they will be using computers that are exponentially more powerful than those in the 1970s; and they may be using a wide variety of tools, machines, and processes. The variety of tools and machines leads to a combinatorial explosion of possible workflows. In addition, tools are based on boundary representations, which are fragile and can easily describe nonsensical objects. This thesis addresses these issues with a set of tools for end-to-end digital fabrication based on volumetric solid models. Workflows are modular, making it easy to add new machines, and a shared core of path-planning operations reduces system complexity. Replacing boundary representations with volumetric representations guarantees that models represent reasonable real-world solids. Adaptively sampled distance fields are used as a generic interchange format. Functional representations are used as a design representation, and we examine scaling behavior and efficient rendering. We present interactive design tools that use these representations as their geometry engine. Data from CT scans is also used to populate these distance fields, showing significant benefits in file size and resolution compared to meshes. Finally, these representations are used as inputs to a modular multimachine CAM workflow. Toolpath generation is implemented, characterized, and tested on a complex solid model. We conclude with a summary of results and recommendations for future research directions.
by Matthew Keeter.
S.M.
10
Miflah, Hussain Ismail Ahamed. « Higher-level representations of natural images ». Thesis, Queen Mary, University of London, 2018. http://qmro.qmul.ac.uk/xmlui/handle/123456789/39759.
Texte intégralRésumé :
The traditional view of vision is that neurons in early cortical areas process information about simple features (e.g. orientation and spatial frequency) in small, spatially localised regions of visual space (the neuron's receptive field). This piecemeal information is then fed-forward into later stages of the visual system where it gets combined to form coherent and meaningful global (higher-level) representations. The overall aim of this thesis is to examine and quantify this higher level processing; how we encode global features in natural images and to understand the extent to which our perception of these global representations is determined by the local features within images. Using the tilt after-effect as a tool, the first chapter examined the processing of a low level, local feature and found that the orientation of a sinusoidal grating could be encoded in both a retinally and spatially non-specific manner. Chapter 2 then examined these tilt aftereffects to the global orientation of the image (i.e., uprightness). We found that image uprightness was also encoded in a retinally / spatially non-specific manner, but that this global property could be processed largely independently of its local orientation content. Chapter 3 investigated if our increased sensitivity to cardinal (vertical and horizontal) structures compared to inter-cardinal (45° and 135° clockwise of vertical) structures, influenced classification of unambiguous natural images. Participants required relatively less contrast to classify images when they retained near-cardinal as compared to near-inter-cardinal structures. Finally, in chapter 4, we examined category classification when images were ambiguous. Observers were biased to classify ambiguous images, created by combining structures from two distinct image categories, as carpentered (e.g., a house). This could not be explained by differences in sensitivity to local structures and is most likely the result of our long-term exposure to city views. Overall, these results show that higher-level representations are not fully dependent on the lower level features within an image. Furthermore, our knowledge about the environment influences the extent to which we use local features to rapidly identify an image.
11
On, Vu Ngoc Minh. « A new minimum barrier distance for multivariate images with applications to salient object detection, shortest path finding, and segmentation ». Electronic Thesis or Diss., Sorbonne université, 2020. http://www.theses.fr/2020SORUS454.
Texte intégralRésumé :
Les représentations hiérarchiques d’images sont largement utilisées dans le traitement d’images pour modéliser le contenu d’une image par un arbre. Une hiérarchie bien connue est l’arbre des formes (AdF) qui encode la relation d’inclusion entre les composants connectés à partir de différents niveaux de seuil. Ce genre d’arbre est auto-duale et invariant de changement de contraste, ce qu’il est utilisé dans de nombreuses applications de vision par ordinateur. En raison de ses propriétés, dans cette thèse, nous utilisons cette représentation pour calculer la nouvelle distance qui appartient au domaine de la morphologie mathématique. Les transformations de distance et les cartes de saillance qu’elles induisent sont généralement utilisées dans le traitement d’images, la vision par ordinateur et la reconnaissance de formes. L’une des transformations de distance les plus couramment utilisées est celle géodésique. Malheureusement, cette distance n’obtient pas toujours des résultats satisfaisants sur des images bruyantes ou floues. Récemment, une nouvelle pseudo-distance, appelée distance de barrière minimale (MBD), plus robuste aux variations de pixels, a été introduite. Quelques années plus tard, Géraud et al. ont proposé une bonne approximation rapide de cette distance : la pseudodistance de Dahu. Puisque cette distance a été initialement développée pour les images en niveaux de gris, nous proposons ici une extension de cette transformation aux images multivariées ; nous l’appelons vectorielle Dahu pseudo-distance. Cette nouvelle distance est facilement et efficacement calculée grâce à à l’arbre multivarié des formes (AdFM). Nous vous proposons une méthode de calcul efficace cette distance et sa carte de saillants déduits dans cette thèse. Nous enquêtons également sur le propriétés de cette distance dans le traitement du bruit et du flou dans l’image. Cette distance s’est avéré robuste pour les pixels invariants. Pour valider cette nouvelle distance, nous fournissons des repères démontrant à quel point la pseudo-distance vectorielle de Dahu est plus robuste et compétitive par rapport aux autres distances basées sur le MB. Cette distance est prometteuse pour la détection des objets saillants, la recherche du chemin le plus court et la segmentation des objets. De plus, nous appliquons cette distance pour détecter le document dans les vidéos. Notre méthode est une approche régionale qui s’appuie sur le saillance visuelle déduite de la pseudo-distance de Dahu. Nous montrons que la performance de notre méthode est compétitive par rapport aux méthodes de pointe de l’ensemble de données du concours Smartdoc 2015 ICDARHierarchical image representations are widely used in image processing to model the content of an image in the multi-scale structure. A well-known hierarchical representation is the tree of shapes (ToS) which encodes the inclusion relationship between connected components from different thresholded levels. This kind of tree is self-dual, contrast-change invariant and popular in computer vision community. Typically, in our work, we use this representation to compute the new distance which belongs to the mathematical morphology domain. Distance transforms and the saliency maps they induce are generally used in image processing, computer vision, and pattern recognition. One of the most commonly used distance transforms is the geodesic one. Unfortunately, this distance does not always achieve satisfying results on noisy or blurred images. Recently, a new pseudo-distance, called the minimum barrier distance (MBD), more robust to pixel fluctuation, has been introduced. Some years after, Géraud et al. have proposed a good and fast-to-compute approximation of this distance: the Dahu pseudodistance. Since this distance was initially developed for grayscale images, we propose here an extension of this transform to multivariate images; we call it vectorial Dahu pseudo-distance. This new distance is easily and efficiently computed thanks to the multivariate tree of shapes (MToS). We propose an efficient way to compute this distance and its deduced saliency map in this thesis. We also investigate the properties of this distance in dealing with noise and blur in the image. This distance has been proved to be robust for pixel invariant. To validate this new distance, we provide benchmarks demonstrating how the vectorial Dahu pseudo-distance is more robust and competitive compared to other MB-based distances. This distance is promising for salient object detection, shortest path finding, and object segmentation. Moreover, we apply this distance to detect the document in videos. Our method is a region-based approach which relies on visual saliency deduced from the Dahu pseudo-distance. We show that the performance of our method is competitive with state-of-the-art methods on the ICDAR Smartdoc 2015 Competition dataset
12
Yang, Yang. « Learning Hierarchical Representations for Video Analysis Using Deep Learning ». Doctoral diss., University of Central Florida, 2013. http://digital.library.ucf.edu/cdm/ref/collection/ETD/id/5892.
Texte intégralRésumé :
With the exponential growth of the digital data, video content analysis (e.g., action, event recognition) has been drawing increasing attention from computer vision researchers. Effective modeling of the objects, scenes, and motions is critical for visual understanding. Recently there has been a growing interest in the bio-inspired deep learning models, which has shown impressive results in speech and object recognition. The deep learning models are formed by the composition of multiple non-linear transformations of the data, with the goal of yielding more abstract and ultimately more useful representations. The advantages of the deep models are three fold: 1) They learn the features directly from the raw signal in contrast to the hand-designed features. 2) The learning can be unsupervised, which is suitable for large data where labeling all the data is expensive and unpractical. 3) They learn a hierarchy of features one level at a time and the layerwise stacking of feature extraction, this often yields better representations. However, not many deep learning models have been proposed to solve the problems in video analysis, especially videos ``in a wild''. Most of them are either dealing with simple datasets, or limited to the low-level local spatial-temporal feature descriptors for action recognition. Moreover, as the learning algorithms are unsupervised, the learned features preserve generative properties rather than the discriminative ones which are more favorable in the classification tasks. In this context, the thesis makes two major contributions. First, we propose several formulations and extensions of deep learning methods which learn hierarchical representations for three challenging video analysis tasks, including complex event recognition, object detection in videos and measuring action similarity. The proposed methods are extensively demonstrated for each work on the state-of-the-art challenging datasets. Besides learning the low-level local features, higher level representations are further designed to be learned in the context of applications. The data-driven concept representations and sparse representation of the events are learned for complex event recognition; the representations for object body parts and structures are learned for object detection in videos; and the relational motion features and similarity metrics between video pairs are learned simultaneously for action verification. Second, in order to learn discriminative and compact features, we propose a new feature learning method using a deep neural network based on auto encoders. It differs from the existing unsupervised feature learning methods in two ways: first it optimizes both discriminative and generative properties of the features simultaneously, which gives our features a better discriminative ability. Second, our learned features are more compact, while the unsupervised feature learning methods usually learn a redundant set of over-complete features. Extensive experiments with quantitative and qualitative results on the tasks of human detection and action verification demonstrate the superiority of our proposed models.Ph.D.
Doctorate
Electrical Engineering and Computer Science
Engineering and Computer Science
Electrical Engineering
13
Vij, Madhav. « Hierarchical lattice vector quantisation of wavelet transformed images ». Thesis, University of Cambridge, 1999. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.624538.
Texte intégral
14
Whyte, Sophie. « Symmetric generation : permutation images and irreducible monomial representations ». Thesis, University of Birmingham, 2006. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.435413.
Texte intégralRésumé :
Symmetric generation has provided concise ways of constructing many classical and sporadic groups, in fact every non-abelian finite simple groups arises in this manner. A symmetric presentation for a group is a homomorphism from a progenitor, p*m : N. onto the group. We give details of our program which constructs all permutation images of a given progenitor. In a monomial progenitor, p*m :m N, the control subgroup N has a monomial action on the symmetric generators of order p > 3. We study monomial progenitors in which the control subgroup has an irreducible monomial representation, as several such progenitors have beell found to map onto sporadic groups. We classify all irreducible monomial representations of the alternating, symmetric and sporadic groups and their covering groups. We use the irreducible monomial representations of the covers of the alternating groups to construct monomial progenitors and we obtain sporadic images of several of these progenitors.
15
Meyer, Christina. « War & ; trauma images in Vietnam War representations ». Hildesheim Zürich New York, NY Olms, 2007. http://d-nb.info/991472861/04.
Texte intégral
16
Ulrich, Markus. « Hierarchical real-time recognition of compound objects in images ». München : Beck, 2003. http://www.loc.gov/catdir/toc/fy0607/2004457892.html.
Texte intégral
17
Nicolov, N. N. « Approximate text generation from non-hierarchical representations in a declarative framework ». Thesis, University of Edinburgh, 1999. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.659996.
Texte intégralRésumé :
This thesis is on Natural Language Generation. It describes a linguistic realisation system that translates the semantic information encoded in a conceptual graph into an English language sentence. The use of a non-hierarchically structured semantic representation (conceptual graphs) and an approximate matching between semantic structures allows us to investigate a more general version of the sentence generation problem where one is not pre-committed to a choice of the syntactically prominent elements in the initial semantics. We show clearly how the semantic structure is declaratively related to linguistically motivated syntactic representation - we use D-Tree Grammars which stem from work on Tree-Adjoining Grammars. The declarative specification of the mapping between semantics and syntax allows for different processing strategies to be exploited. A number of generation strategies have been considered: a pure top-down strategy and a chart-based generation technique which allows partially successful computations to be reused in other branches of the search space. Having a generator with increased paraphrasing power as a consequence of using non-hierarchical input and approximate matching raises the issue whether certain 'better' paraphrases can be generated before others. We investigate preference-based processing in the context of generation.
18
Timoner, Samson J. (Samson Joshua) 1975. « Compact representations for fast nonrigid registration of medical images ». Thesis, Massachusetts Institute of Technology, 2003. http://hdl.handle.net/1721.1/29950.
Texte intégralRésumé :
Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2003.Includes bibliographical references (p. 167-179).
We develop efficient techniques for the non-rigid registration of medical images by using representations that adapt to the anatomy found in such images. Images of anatomical structures typically have uniform intensity interiors and smooth boundaries. We create methods to represent such regions compactly using tetrahedra. Unlike voxel-based representations, tetrahedra can accurately describe the expected smooth surfaces of medical objects. Furthermore, the interior of such objects can be represented using a small number of tetrahedra. Rather than describing a medical object using tens of thousands of voxels, our representations generally contain only a few thousand elements. Tetrahedra facilitate the creation of efficient non-rigid registration algorithms based on finite element methods (FEM). We create a fast, FEM-based method to non-rigidly register segmented anatomical structures from two subjects. Using our compact tetrahedral representations, this method generally requires less than one minute of processing time on a desktop PC. We also create a novel method for the non-rigid registration of gray scale images. To facilitate a fast method, we create a tetrahedral representation of a displacement field that automatically adapts to both the anatomy in an image and to the displacement field. The resulting algorithm has a computational cost that is dominated by the number of nodes in the mesh (about 10,000), rather than the number of voxels in an image (nearly 10,000,000). For many non-rigid registration problems, we can find a transformation from one image to another in five minutes. This speed is important as it allows use of the algorithm during surgery.
(cont.) We apply our algorithms to find correlations between the shape of anatomical structures and the presence of schizophrenia. We show that a study based on our representations outperforms studies based on other representations. We also use the results of our non-rigid registration algorithm as the basis of a segmentation algorithm. That algorithm also outperforms other methods in our tests, producing smoother segmentations and more accurately reproducing manual segmentations.
by Samson J. Timoner.
Ph.D.
19
CHATER, ABDELHAFID. « Images et representations maitre-eleve a l'ecole primaire marocaine ». Université Marc Bloch (Strasbourg) (1971-2008), 1992. http://www.theses.fr/1992STR20015.
Texte intégralRésumé :
Bien que notre etude se situe dans le cadre d'une contribution generale a l'etude maitre-eleve, nous avons envisage plus particulierement d'etudier l'image qu'a l'eleve de son maitre et la representation que celui-ci se fait de cette image. Dans le domaine de l'image du maitre chez l'eleve, nous avons voulu savoir si l'enfant marocain accorde plus d'importance aux qualites humaines et relationnelles de l'enseignant plutot qu'a ses qualites relatives a la technique pedagogique. Le but pratique de l'enquete aupres des eleves etait donc d'analyser les images qu'ils possedent de leur maitre, d'etudier toutes les variables possibles qui influencent la construction de cette image et enfin de definir les attentes qui en decoulent. Quant a notre enquete aupres des maitres, elle a autant porte sur les attitudes, attentes et conceptions de la relation maitre-eleve qu'a leurs representations du role de l'enseignant, dans la mesure ou la notion de representation est multidimensionnelle. En effet, cette derniere enquete a eu deux objectifs : le premier etait de reperer les types des relations auxquelles adherent les enseignants dans leurs discours. Le deuxieme etait de preciser l'image qu'ils se font de leur personnageEven through the present study can be understood within a broad 9contribution framework related to the study of the teacher-pupil relation ship, we have especialy considered both the pupil's image of his teacher and this latter's personal representation of that image. As perceived by pupil, we have tried to know if the moroccah pupil gives more importance to the human and relational qualities the practical purpose of the investigation has been to analyze the pupil's image of their teacher, to study any possible variables which may influence the elaboration of that image and finally define the foregone expectations. As to our inquiry on teachers, it concerns attitudes, expectations of the teacher-pupil relationship as well as their representations pf the teacher's role given the multidimensionality of representation as a notion. By and large, the inquiry on teachers has two aims; the first consisted in pinning down the kind of relations to which the teacher's discourse is committed the second has been to specify the image that both have of themselves
20
Høvik, Ingeborg. « Arctic images 1818-1859 ». Thesis, University of Edinburgh, 2013. http://hdl.handle.net/1842/12261.
Texte intégralRésumé :
This thesis asks whether there existed a unified view of the Arctic during the time period connected to the high point of British endeavour to find a Northwest Passage, from the first expeditions of the nineteenth-century in 1818 to the return of the last Franklin search party in 1859, forty-one years later. Using this time frame as its marker, the focus of the thesis is primarily on British representations of Arctic landscapes, exploration and Inuit peoples. Through careful empirical analysis of a variety of media, including professional painting, on-the-spot sketches, prints and popular exhibitions, it examines from an art historical viewpoint the historical, political, social and aesthetic contexts in which Arctic representations occurred.
21
Cosmas, John Paul. « The generation of 3-dimensional object representations from range images ». Thesis, Imperial College London, 1988. http://hdl.handle.net/10044/1/47007.
Texte intégral
22
Cichocki, Radoslaw. « Classification of objects in images based on various object representations ». Thesis, Blekinge Tekniska Högskola, Avdelningen för programvarusystem, 2006. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-5774.
Texte intégralRésumé :
Object recognition is a hugely researched domain that employs methods derived from mathematics, physics and biology. This thesis combines the approaches for object classification that base on two features – color and shape. Color is represented by color histograms and shape by skeletal graphs. Four hybrids are proposed which combine those approaches in different manners and the hybrids are then tested to find out which of them gives best results.Mail the author at radoslaw.cichocki(at)gmail.com
23
Penn, Gemma Louise. « Medicalization and representations of smoking in public discourse and images ». Thesis, London School of Economics and Political Science (University of London), 1998. http://etheses.lse.ac.uk/1503/.
Texte intégralRésumé :
An approach to smoking through an analysis of its representations, grounded in the medicalization literature, highlights the inadequacies of a narrow medical perspective and some of its negative implications. This does not require that we abandon the medical discourse, but stresses the importance of setting it in a wider representational context. Drawing on the work of Saussure, Barthes, Eco and Foucault, the author constructs a theory of interaction amongst representations suited to both discourse and images. To investigate the medicalization of smoking, four empirical studies are reported which include quantitative and qualitative approaches to press reporting at both macro-and micro-levels, cigarette advertising and packaging. There have been medical representations of smoking since the introduction of tobacco into Britain. However, a thematic analysis of tobacco-related reporting in the Times newspaper (1946-1995) found that these representations have expanded and diversified, becoming increasingly linked to other representations (e.g. financial) and generating new themes (e.g. discrimination, litigation). Medical representations, however, are contested and subject to subversion by alternative representations, including libertarian and alternative medical constructions of smoking. These processes are investigated in a detailed structural and rhetorical analysis of a contemporary newspaper article, together with related correspondence and cartoons. Quantitative and qualitative analyses of 754 UK cigarette advertisements from four national newspapers (1946-1995) identified an increase in medical and packaging-related representations of smoking and a decrease in financial representations, representations of the act of smoking and of cigarettes as social currency. The final, questionnaire, study (with 60 participants) found, among other things, a clear and consensually-held system of health-related signification in contemporary UK cigarette packaging in ratings of packets. The thesis concludes with a discussion of the implications of medicalization for smoking-related policy and for the smoker, and of smoking for the medicalization literature.
24
Chongburee, Wachira. « Implementation of Iterative Reconstruction of Images from Multiple Bases Representations ». Thesis, Virginia Tech, 1998. http://hdl.handle.net/10919/35379.
Texte intégralRésumé :
Usually, image compression techniques that use only one transform exhibit some poor properties. For instance, the Discrete Cosine Transform (DCT) cannot efficiently represent high frequency components, resulting in blurred images. The Multiple Bases Representation (MBR) compression technique, which uses two or more transforms, is found to be superior to the single transform techniques in terms of representation efficiency. However, some bits in the MBR representation are needed to track the basis information. The MBR image quality is deteriorated by discontinuities at block boundaries, as is the standard DCT transform.
In this thesis, test images are distorted by MBR compression using a Recursive Residual Projection algorithm. This algorithm is a sub-optimal method to find the best basis vector subset for representing images based on multiple orthogonal bases. The MBR distorted images are reconstructed by the iterative method of Projection onto Convex Sets (POCS). Many constraints that form convex sets are reviewed and examined.
Due to the high distortion at the block boundaries, some constraints are introduced particularly to reduce artifacts at the boundaries. Some constraints add energy to the reconstructed images while others remove energy. Thus, the initial vectors play a key role in the performance of the POCS method for better MBR reconstruction. This thesis also determines the most appropriate initial vector for each constraint.
Finally, the composite projections associated with the sign, minimum decreasing and norm-of-slope constraints are used to improve the reconstruction of the MBR distorted images and the effect of ordering of the projections is investigated.
Master of Science
25
Truong, Kwan K. « Vector quantizer design for images and video based on hierarchical structures ». Diss., Georgia Institute of Technology, 1992. http://hdl.handle.net/1853/13266.
Texte intégral
26
Selvan, A. D. Arul Nirai. « Highlighting dissimilarity in medical images using hierarchical clustering based segmentation (HCS) ». Thesis, Sheffield Hallam University, 2007. http://shura.shu.ac.uk/20342/.
Texte intégralRésumé :
Tissue abnormality in a medical image is usually related to a dissimilar part of an otherwise homogeneous image. The dissimilarity may be subtle or strong depending on the medical modality and the type of abnormal tissue. Dissimilarity within an otherwise homogeneous area of an image may not always be due to tissue abnormality. It might be due to image noise or due to variability within the same tissue type. Given this situation it is almost impossible to design and implement a generic segmentation process that will consistently give a single appropriate solution under all conditions. Hence a dissimilarity highlighting process that yields a hierarchy of segmentation results is more useful. This would benefit from high level human interaction to select the appropriate image segmentation for a particular application, because one of the capabilities of the human vision process when visualising images is its ability to visualise them at different levels of details. The purpose of this thesis is to design and implement a segmentation procedure to resemble the capability of the human vision system's ability to generate multiple solutions of varying resolutions. To this end, the main objectives for this study were: (i) to design a segmentation process that would be unsupervised and completely data driven. (ii) to design a segmentation process that would automatically and consistently generate a hierarchy of segmentation results. In order to achieve these objectives a hierarchical clustering based segmentation (HCS) process was designed and implemented. The developed HCS process partitioned the images into their constituent regions at hierarchical levels of allowable dissimilarity between the different spatially adjacent or disjoint regions. At any particular level in the hierarchy the segmentation process clustered together all the pixels and/or regions that had dissimilarity among them which was less than or equal to the dissimilarity allowed for that level. The clustering process was designed in such a way that the merging of the clusters did not depend on the order in which the clusters were evaluated. The HCS process developed was used to process images of different medical modalities and the results obtained are summarised below: (i) It was successfully used to highlight hard to visualise stroke affected areas in T2 weighted MR images confirmed by the diffusion weighted scans of the same areas of the brain. (ii) It was used to highlight dissimilarities in the MRI, CT and ultrasound images and the results were validated by the radiologists. It processed medical image data and consistently produced a hierarchy of segmentation results but did not give a diagnosis. This was left for the experts to make use of the results and incorporate these with their own knowledge to arrive upon a diagnosis. Thus the process acts as an effective computer aided detection (CAD) tool. The unique features of the designed and implemented HCS process are: (i) The segmentation process is unsupervised, completely data driven and can be applied to any medical modality, with equal success, without any prior information about the image data(ii) The merging routines can evaluate and merge spatially adjacent and disjoint similar regions and consistently give a hierarchy of segmentation results. (iii) The designed merging process can yield crisp border delineation between the regions.
27
Westwell, Guy. « History-in-images/images in history : American cultural memory and film representations of the Vietnam War ». Thesis, University of Glasgow, 2001. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.340278.
Texte intégralRésumé :
This thesis charts points of convergence between the fields of historical studies and film studies that generate a line of inquiry which questions how the development and dissemination of film and television have significantly shaped historical conscIOusness. Taking this line of inquiry as a starting point, this thesis identifies the ways in which film (and television) representations have informed American cultural memory of the Vietnam War. The thesis describes how the reporting of the war in newspapers and on television results in the production of a number of vivid and powerful 'nodal images'; these images enable their viewers to locate themselves in relation to the larger event and offer guidance regarding how other representations produced in response to the war might be understood. The thesis goes on to explore how these images play a significant. role in secondary film and television representations, including Hollywood feature films, whereby the initial connotations of the image are recirculated, reenacted and re-scripted. The thesis also indicates how other film representation of the war - such as the film records produced by the American military for tactical and strategic purposes and amateur film produced by American military personnel- are side-lined by the dominance of these nodal images. This study closes by proposing a taxonomy of the key features of these film (and television) representations and profiles the ways in which these features determi~e American cultural memory of the war and mediate historical experience more generally. The conclusion arrived at is that the historical consciousness engendered by these representations encourages the meaning of the Vietnam War to be located in relation to individual phenomenological experience and that the priVileging of this experience above all others marginalises the wider frames of reference - politics, history, economics and so on - which might make that experience meaningful.
28
Hilliard, Kristina Marie. « Images of a Gendered Kingship : Visual Representations of Hatshepsut and Her Influence on Images of Nefertiti ». Thesis, University of North Texas, 2006. https://digital.library.unt.edu/ark:/67531/metadc5323/.
Texte intégralRésumé :
I investigate why gendered images of Hatshepsut influenced androgynous images of Nefertiti in New Kingdom Egypt and how Nefertiti and Akhenaten used their images in the promotion of their monotheistic religion; through a contextual, stylistic and feminist examination of the images. Hatshepsut cultivated images of herself to legitimize her rule in relation to canonical kings before her. Similarly, Nefertiti represented herself as a figure indiscernible from Akhenaten, creating an image of female co-rulership. Although the visual representations of both Hatshepsut and Nefertiti differ, the concepts behind each are analogous. They both manipulated androgyny to create images displaying powerful women equal in status to male Egyptian kings.
29
Farber, Lee. « Symmetric generation of finite homomorphic images ? » CSUSB ScholarWorks, 2005. https://scholarworks.lib.csusb.edu/etd-project/2901.
Texte intégralRésumé :
The purpose of this thesis was to present the technique of double coset enumeration and apply it to construct finite homomorphic images of infinite semidirect products. Several important homomorphic images include the classical groups, the Projective Special Linear group and the Derived Chevalley group were constructed.
30
Goh, Hanlin. « Learning deep visual representations ». Paris 6, 2013. http://www.theses.fr/2013PA066356.
Texte intégralRésumé :
Les avancées récentes en apprentissage profond et en traitement d'image présentent l'opportunité d'unifier ces deux champs de recherche complémentaires pour une meilleure résolution du problème de classification d'images dans des catégories sémantiques. L'apprentissage profond apporte au traitement d'image le pouvoir de représentation nécessaire à l'amélioration des performances des méthodes de classification d'images. Cette thèse propose de nouvelles méthodes d'apprentissage de représentations visuelles profondes pour la résolution de cette tache. L'apprentissage profond a été abordé sous deux angles. D'abord nous nous sommes intéressés à l'apprentissage non supervisé de représentations latentes ayant certaines propriétés à partir de données en entrée. Il s'agit ici d'intégrer une connaissance à priori, à travers un terme de régularisation, dans l'apprentissage d'une machine de Boltzmann restreinte (RBM). Nous proposons plusieurs formes de régularisation qui induisent différentes propriétés telles que la parcimonie, la sélectivité et l'organisation en structure topographique. Le second aspect consiste au passage graduel de l'apprentissage non supervisé à l'apprentissage supervisé de réseaux profonds. Ce but est réalisé par l'introduction sous forme de supervision, d'une information relative à la catégorie sémantique. Deux nouvelles méthodes sont proposées. Le premier est basé sur une régularisation top-down de réseaux de croyance profonds à base de RBMs. Le second optimise un cout intégrant un critre de reconstruction et un critre de supervision pour l'entrainement d'autoencodeurs profonds. Les méthodes proposées ont été appliquées au problme de classification d'images. Nous avons adopté le modèle sac-de-mots comme modèle de base parce qu'il offre d'importantes possibilités grâce à l'utilisation de descripteurs locaux robustes et de pooling par pyramides spatiales qui prennent en compte l'information spatiale de l'image. L'apprentissage profonds avec agrÉgation spatiale est utilisé pour apprendre un dictionnaire hiÉrarchique pour l'encodage de reprÉsentations visuelles de niveau intermÉdiaire. Cette mÉthode donne des rÉsultats trs compétitifs en classification de scènes et d'images. Les dictionnaires visuels appris contiennent diverses informations non-redondantes ayant une structure spatiale cohérente. L'inférence est aussi très rapide. Nous avons par la suite optimisé l'étape de pooling sur la base du codage produit par le dictionnaire hiérarchique précédemment appris en introduisant introduit une nouvelle paramétrisation dérivable de l'opération de pooling qui permet un apprentissage par descente de gradient utilisant l'algorithme de rétro-propagation. Ceci est la premire tentative d'unification de l'apprentissage profond et du modèle de sac de mots. Bien que cette fusion puisse sembler évidente, l'union de plusieurs aspects de l'apprentissage profond de représentations visuelles demeure une tache complexe à bien des égards et requiert encore un effort de recherche importantRecent advancements in the areas of deep learning and visual information processing have presented an opportunity to unite both fields. These complementary fields combine to tackle the problem of classifying images into their semantic categories. Deep learning brings learning and representational capabilities to a visual processing model that is adapted for image classification. This thesis addresses problems that lead to the proposal of learning deep visual representations for image classification. The problem of deep learning is tackled on two fronts. The first aspect is the problem of unsupervised learning of latent representations from input data. The main focus is the integration of prior knowledge into the learning of restricted Boltzmann machines (RBM) through regularization. Regularizers are proposed to induce sparsity, selectivity and topographic organization in the coding to improve discrimination and invariance. The second direction introduces the notion of gradually transiting from unsupervised layer-wise learning to supervised deep learning. This is done through the integration of bottom-up information with top-down signals. Two novel implementations supporting this notion are explored. The first method uses top-down regularization to train a deep network of RBMs. The second method combines predictive and reconstructive loss functions to optimize a stack of encoder-decoder networks. The proposed deep learning techniques are applied to tackle the image classification problem. The bag-of-words model is adopted due to its strengths in image modeling through the use of local image descriptors and spatial pooling schemes. Deep learning with spatial aggregation is used to learn a hierarchical visual dictionary for encoding the image descriptors into mid-level representations. This method achieves leading image classification performances for object and scene images. The learned dictionaries are diverse and non-redundant. The speed of inference is also high. From this, a further optimization is performed for the subsequent pooling step. This is done by introducing a differentiable pooling parameterization and applying the error backpropagation algorithm. This thesis represents one of the first attempts to synthesize deep learning and the bag-of-words model. This union results in many challenging research problems, leaving much room for further study in this area
31
Amorós, Carafí Laia. « Images of Galois representations and p-adic models of Shimura curves ». Doctoral thesis, Universitat de Barcelona, 2016. http://hdl.handle.net/10803/471452.
Texte intégralRésumé :
The Langlands program is a vast and unifying network of conjectures that connect the world of automorphic representations of reductive algebraic groups and the world of Galois representations.
These conjectures associate an automorphic representation of a reductive algebraic group to every n-dimensional representation of a Galois group, and the other way around: they attach a Galois representation to any automorphic representation of a reductive algebraic group. Moreover, these correspondences are done in such a way that the automorphic L-functions attached to the two objects coincide.
The theory of modular forms is a field of complex analysis whose main importance lies on its connections and applications to number theory. We will make use, on the one hand, of the arithmetic properties of modular forms to study certain Galois representations and their number theoretic meaning. On the other hand, we will use the geometric meaning of these complex analytic functions to study a natural generalization of modular curves. A modular curve is a geometric object that parametrizes isomorphism classes of elliptic curves together with some additional structure depending on some modular subgroup. The generalization that we will be interested in are the so called Shimura curves. We will be particularly interested in their p-adic models.
In this thesis, we treat two different topics, one in each side of the Langlands program. In the Galois representations' side, we are interested in Galois representations that take values in local Hecke algebras attached to modular forms over finite fields. In the automorphic forms' side, we are interested in Shimura curves: we develop some arithmetic results in definite quaternion algebras and give some results about Mumford curves covering p-adic Shimura curves.
32
Abraham, Christiana. « Images of third world women : difference and disjuncture in development representations ». Thesis, McGill University, 2008. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=18780.
Texte intégralRésumé :
Images of Third World Women: Difference and Disjuncture in Development Representations This dissertation analyzes the use of images of women of the third world by the Canadian International Development Agency (CIDA) during the Women in Development period (WID). This critical interpretative analysis examines how images as institutional texts function to reproduce notions of difference about women and the third world. The study looks at the disjuncture between the language and images of development and studies how media narratives circulate gender discourses through strands of colonial and race-based knowledges as normalized ways of classifying the world through imaginary categories. The study suggests that through recurrent representative tropes, WID facilitated the deliberate production of a development category that was highly identifiable – that of the third world woman. This focus on women allowed their images to assume immense circulatory value in CIDA development imagery so that depictions of third world women became representative of development itself. Operating as sign systems, these images of third world women also functioned as important symbols of the uniqueness of Canadian values. The study provides a framework for understanding development representations as powerful sign systems that prioritize and hierarchize preferred narratives.Résumé Les Images des Femmes du Tiers Monde: Différence et Rupture dans les Représentations du Développement Cette thèse analyse l'usage des images des femmes du Tiers Monde par l'Agence Canadienne de Développement International (ACDI) durant la période 'Les Femmes dans le Développement' (FED). Cette analyse interprétative critique examine comment ces images, - en tant que textes institutionnels -, fonctionnent pour reproduire les notions de différence à propos des femmes et du Tiers Monde. Cette étude se penche sur la rupture entre le langage et les images du développement et analyse la manière dont les messages médiatiques véhiculent les discours relativement au genre par le truchement de formes résiduelles de savoirs colonial et racisé en tant que moyens normalisés de classifier le monde par l'entremise de catégories imaginaires. Cette étude suggère que, par l'entremise de catégories représentatives récurrentes, l'époque FED facilita la production délibérée d'une figure du développement fortement identifiable - celle de la femme du Tiers Monde. Cette tendance à cibler les femmes a permis à ces images d'assurer une immense valeur véhiculaire dans l'imaginaire du développement de l'ACDI, de sorte que ces images des femmes du Tiers Monde sont devenues elles-mêmes représentatives du développement. Fonctionnant comme systèmes de signes, ces images des femmes du Tiers Monde ont également opérés comme d'importants symboles de la singularité des valeurs canadiennes. Cette étude fournit un cadre pour comprendre les représentations du développement en tant que systèmes de signes puissants qui priorisent et hiérarchisent les préférences narratives.
33
Lindohf, Jessica Malin. « Images of the end : representations of the apocalyptic in contemporary film ». Thesis, University of Glasgow, 2002. http://theses.gla.ac.uk/3104/.
Texte intégralRésumé :
This thesis sets out to investigate the relationship between the ‘classical apocalypse’ and the contemporary apocalypse as portrayed by the films A Clockwork Orange (1971), Apocalypse Now (1979) and Crash (1995). The ‘classical apocalypse’ is a literary genre which supplies a rich and vivid imagery where the image takes precedence over the narrative. At the centre of the ‘classical apocalypse’ is the image, and this thesis explores it the imagery of apocalypse can be translated from its traditional literary form to the visual form of film. The apocalypse is a revealing of that which has been concealed and which lies in the future of humankind at the end of time. In the postmodern era with the absence of meaning, apocalypse and God, the apocalypse has become a nihilistic repetition and the revealing has become feared since it might be a revealing of nothing. These contemporary depictions of the end, I would argue, help the apocalypse to come into its own in a postmodern setting, and the medium of film offers a possibility to further emphasise the visuality and potent imagery of the end, expressing the concerns of the apocalypse fully. As such they provide a ‘sense of an ending’ and an apocalyptic sentiment which is an unnerving and evasive as the ‘classical apocalypse’. These films revisit as well as revamp and rehearse the imagery of the Biblical apocalypse, becoming a-theological statements if not on the Bible, on the state of society and the apocalypse.
34
Fenwick, Melissa E. « Reel Images : Representations of Adult Male Prisons by the Film Industry ». Scholar Commons, 2009. https://scholarcommons.usf.edu/etd/1962.
Texte intégralRésumé :
Research on the criminal justice system, punishment, and media continue to generate academic interest, particularly in the realm of social constructionism. The social construction perspective provides insight into the process through which media-controlled images are translated into social definitions of crime and justice. One new area of interest is the representations of prisons and penal culture by the entertainment media, namely the film industry. In this study, the author contributes to the area of social constructionist literature by administering a content analysis of eleven feature films on male prisons produced between 1979 and 2001. The author examines the frequency and context of several constructs of penal culture: drug use and trafficking, rape and sexual assault, violence, and gang affiliation. This research examines whether the representations of these issues in recent motion pictures are consistent with extant academic correctional literature.
The present study found that within prison films the amount of portrayal of drug use and trafficking, and rape and sexual assault is consistent with the academic literature. Overall, when compared to the academic literature, prison movies under represent gang affiliation but within movies that portray gang affiliation, that portrayal is similar to the academic literature. Notably, heroin was the drug of choice depicted within prison films while academic correctional research in prisons shows marijuana as the drug of choice. The most significant finding was that the amount and type of violence, specifically murder, was overrepresented in prison films compared to the amount and type of violence reported within current academic research.
The over emphasis on violence and killing within prison films and the representation of heroin as the major drug consumed and trafficked could lead to public misunderstanding about the realities of prison life and living conditions of the prison institution. This study provides not only noteworthy information concerning the representations of prison life and penal culture by the film industry but also insight into the inconsistencies between the information presented on film and that within academic correctional literature that are transferred via this medium to the general public.
35
Schick, Wiebke [Verfasser], et Hanspeter A. [Akademischer Betreuer] Mallot. « Acquisition and consolidation of hierarchical representations of space / Wiebke Schick ; Betreuer : Hanspeter A. Mallot ». Tübingen : Universitätsbibliothek Tübingen, 2018. http://d-nb.info/1168634679/34.
Texte intégral
36
Tzanoudaki, Stephany. « The transition of the Olympic city from visual representations Of Coubertin's modern ideal to city representations as fashionable images ». Thesis, University of Edinburgh, 2011. http://hdl.handle.net/1842/9380.
Texte intégralRésumé :
My interest in this particular subject, involving the analysis of design and architectural proposals for the Olympic city, derives from an earlier MPhil study, completed in 2001 (see section 3.1.2). My PhD research gave me the chance to further my investigation in this subject and focus on the role of Olympic design and architecture and the analysis of the material that describes the aesthetic character and language of host cities. By writing this thesis, I have connected, the Olympic city subject with critical theory involving issues of modern cities’ changing urban, social and cultural identities, reflected in their represented image, and I have also used methods of visual analysis. Within the course of the last ten years of my research in this field, I realised that there is a research gap in the investigation of the visual identity of different Olympic cities. This research gap consists of a lack of research that attempts a critical review of the role of Olympic architecture and design since the first Summer Olympic Games in Athens 1896 and also a lack of examination and critical analysis of the visual (and descriptive) material available in the Olympic cities’ archives. More specifically, this analysis involves questions about: - the production of a ‘visual identity’ (i.e. information about the producer of this identity, the design criteria and influences, such as technological and ideological); - the relationship of this identity with the surrounding urban, social and cultural environment: ‘what knowledge is being deployed and whose knowledge is excluded from this representation’ (Rose, 2007, p.259); 10 - the changes in how this identity is communicated and perceived (changes in the audience, in the perception of the event as a spectacle, in media and in the means of communication involved). My research journey, from the gathering of the necessary material to the writing up of this thesis, has involved a historical review of archives from different sources (libraries, organising committees and the press). It has also involved a review of theories that facilitated the interpretation of the Olympic city development as a visual identity, placed within a historical and sociocultural context. Methodologically, therefore, this study is a synthesis based on both the gathering of secondary data and also critical theory on art, design and architecture and on cities’ urban and social development. The work of the following researchers has been especially useful in exploring many of the thesis’ arguments: theorists such as Walter Benjamin and Ernst Bloch who are concerned with the modern city development; David Harvey, Fredric Jameson, Jean Baudrillard and Zygmunt Bauman, who examine the transformations in twentieth century social and cultural conditions (modernism and postmodernism); and also Sharon Zukin, Kevin Robins, and Anna Klingmann, who examine the social and cultural transformations in contemporary urban development. In this study, I also focus on the idea of transition and, therefore, on Olympic city examples that have made an immense effort to create a visual identity or alter the ways that people visually experienced them. I have been particularly interested in case studies that, with their design and architectural ideas and the ways they visually communicated the identity of a place, contributed to the transition of the Olympic city in its development as a visual identity. I suggest that, in addition to the examination of the Olympic city development as an urban plan (changing in size and scale and engaging with new technologies) and also as an international event (emerging to a mega-event), there is another type of development in Olympic design and architecture that is worth exploring, that of the Olympic city as a visual identity. 11 The original ideas in this thesis have to do with the development of the Olympic city as a visual identity and, therefore, with: - the analysis of visual material (photographs, maps, plans, pictures from the press and posters) and written material (from the Official Olympic reviews, the press, books, letters and speeches) having to do with Olympic cities since the Summer Olympic Games in Athens 1896. - the analysis of this material by interpreting the characteristics (design criteria, and priorities, who takes decisions and who is the image maker) behind each Olympic city’s design and architecture proposals. This analysis considers the promoted urban, social and cultural profile of the host city, but also considers any alternative (different from the represented) urban, social and cultural identity of the host city. - the selection of examples from the Olympic cities' visual identities that best represent the Olympic city as a transition from the modern to the postmodern era, based on characteristics from these visual identities that have faded, altered or been abolished and also characteristics that have been emphasised and promoted. Many of these characteristics changed the contemporary shape and represented profile not only of Olympic cities but also of cities in general. - the critical analysis of the role of design and architecture in the representation of an Olympic city, reflected in the characteristics of its visual identity.
37
El-Harby, Ahmed Ahmed Abd El-Fattah. « Automatic extraction of vector representations of line features from remotely sensed images ». Thesis, Keele University, 2001. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.344096.
Texte intégral
38
Waldroup, Heather Leigh. « Traveling images : representations of the south pacific from colonial and postcolonial worlds / ». Diss., Digital Dissertations Database. Restricted to UC campuses, 2004. http://uclibs.org/PID/11984.
Texte intégral
39
Leyk, Williams Malgorzata. « Summarizing FLARE assay images in colon carcinogenesis ». Texas A&M University, 2004. http://hdl.handle.net/1969.1/3132.
Texte intégralRésumé :
Intestinal tract cancer is one of the more common cancers in the United States. While in some individuals a genetic component causes the cancer, the rate of cancer in the remainder of the population is believed to be affected by diet. Since cancer usually develops slowly, the amount of oxidative damage to DNA can be used as a cancer biomarker. This dissertation examines effective ways of analyzing FLARE assay data, which quantifies oxidative damage. The statistical methods will be implemented on data from a FLARE assay experiment, which examines cells from the duodenum and the colon to see if there is a difference in the risk of cancer due to corn or fish oil diets. Treatments of the oxidizing agent dextran sodium sulfate (DSS), DSS with a recovery period, as well as a control will also be used.
Previous methods presented in the literature examined the FLARE data by summarizing the DNA damage of each cell with a single number, such as the relative tail moment (RTM). Variable skewness is proposed as an alternative measure, and shown to be as effective as the RTM in detecting diet and treatment differences in the standard analysis. The RTM and skewness data is then analyzed using a hierarchical model, with both the skewness and RTM showing diet/treatment differences. Simulated data for this model is also considered, and shows that a Bayes Factor (BF) for higher dimensional models does not follow guidelines presented by Kass and Raftery (1995).
It is hypothesized that more information is obtained by describing the DNA damage functions, instead of summarizing them with a single number. From each function, seven points are picked. First, they are modeled independently, and only diet effects are found. However, when the correlation between points at the cell and rat level is modeled, much stronger diet and treatment differences are shown both in the colon and the duodenum than for any of the previous methods. These results are also easier to interpret and represent graphically, showing that the latter is an effective method of analyzing the FLARE data.
40
Dunn, David Richard. « Home truths from abroad : television representations of the tourist destination ». Thesis, University of Birmingham, 1998. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.343895.
Texte intégral
41
Xu, Shanna, et Pengzhan You. « Western images of China : media representations of Chinese attempts to invest in Saab ». Thesis, Högskolan Väst, Institutionen för ekonomi och it, 2012. http://urn.kb.se/resolve?urn=urn:nbn:se:hv:diva-4390.
Texte intégralRésumé :
The aim of this thesis is to describe Western images of China by focusing on media representations of Chinese attempts to invest in Saab. Theories of media representation, orientalism, racialization and stereotype are applied and used in the qualitative discourse analysis in order to find out if there are orientalist and racialized stereotypes in the material. The findings show that there are orientalist stereotypes and racialized stereotypes presented in the material. The analysis also sums up that China is a country whose people are represented to be adaptable and to have amazing productivity, since China has cheap labor power and lax labor law. Furthermore, China is represented as a country whose financial power is strong and solid, Western media characterizes China as a threat. Moreover, Chinese negotiators who went to Sweden to negotiate not only are represented as full of ambitions, but also they are seen as the saviors for Western companies which are on the verge of bankrupt. This thesis contributes to the literature by filling the gap about the Chinese attempts to invest in Saab, which is characterized by Western media.
42
Sullivan, Sandra Jean. « Representations of Mary of Modena, Duchess, Queen Consort and exile : images and texts ». Thesis, University College London (University of London), 2008. http://discovery.ucl.ac.uk/1349620/.
Texte intégralRésumé :
This study of the way in which the "image" of Mary of Modena was constructed, and then deconstructed, examines the changing ways in which this late seventeenth-century Italian Catholic aristocrat was represented, both visually in portraits,prints and engravings, and textually in verse and prose, opera, masque and pamphlet plays, after her arrival in England in 1673 as the wife of James, Duke of York (later James II). The images and texts are placed within the cultural context of the age: contemporary artists, writers, patronage, conventions and expectations, including those relating to women, are discussed. The focus of the first chapter is James's first wife, Anne Hyde. A comparison of the manner in which these two women, both Catholic, were represented, one coming from an ancient and noble foreign house,the other from an undistinguished country family, is instructive. The next four chapters examine the themes of "beauty", her "otherness", her patronage of music, artists and writers, and her court, during the years 1673 to 1685 when Mary of Modena was Duchess of York. The contrast between the portraits, music and poetry produced at the time of James II's coronation in 1685 and the representations produced as part of the adverse propaganda campaign waged by James's son-in-law, William of Orange which culminated in the "warming pan scandal" after the birth of Mary of Modena's son ("The Old Pretender"), and led to the "Glorious Revolution" of 1688 are the subject of chapters 6 and 7. The final chapters how show Mary of Modena was represented during her exile in France from 1689 until her death there in 1718.
43
Bülow, Thomas [Verfasser]. « Hypercomplex Spectral Signal Representations for the Processing and Analysis of Images / Thomas Bülow ». Kiel : Universitätsbibliothek Kiel, 1999. http://d-nb.info/1080332537/34.
Texte intégral
44
Frappart, Virginie. « La mise en mots de la ville contemporaine, representations et images de nantes ». Nantes, 2001. http://www.theses.fr/2001NANT3001.
Texte intégralRésumé :
Depuis une vingtaine d'annees, des relocalisations territoriales ainsi qu'une redefinition du pouvoir local sont en jeu a nantes comme dans la plupart des grandes villes francaises. On note egalement une evolution dans la facon de cerner les problemes et les enjeux de la ville. Ces territorialisations des enjeux urbains passent par les differentes representations de la ville et se traduisent notamment par la maniere de nommer et de caracteriser les espaces auxquels on se refere, ce que j'appelle la mise en mots de la ville contemporaine. Plus que la ville de nantes, l'objet de cette recherche porte sur ses representations, car si la ville ne se laisse pas saisir hors de sa materialite, elle n'est produite comme unite qu'au travers du sens qu'on lui attribue. Il convient donc de decrypter les discours sur nantes et d'analyser comment, a partir de ces discours, se degagent des << representations en actes >>. Dans l'hypothese que l'emergence des images urbaines est consubstantielle a la construction de projets territoriaux portes par de nouveaux acteurs urbains, ce sont les rapports dialectiques entre les representations et les systemes d'acteurs locaux qui sont au centre de cet exercice universitaire. En partant du principe que chaque acteur, inscrit dans un groupe social, a une vision, un angle d'attaque pour apprehender la ville mais egalement des objectifs et une pratique specifiques au role et a la place qu'il occupe, il est interessant de confronter les differentes images constitutives d'une representation, ainsi que de degager l'existence d'ideologies urbaines et de strategies territoriales
45
Frey, Nathan J. « Hierarchical Latent Networks for Image and Language Correlation ». Ohio University / OhioLINK, 2011. http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1320847134.
Texte intégral
46
Lynn, Jennifer M. Hagemann Karen. « Contesting images representations of the Modern Woman in the German illustrated press, 1924-1933 / ». Chapel Hill, N.C. : University of North Carolina at Chapel Hill, 2008. http://dc.lib.unc.edu/u?/etd,1536.
Texte intégralRésumé :
Thesis (M.A.)--University of North Carolina at Chapel Hill, 2008.Title from electronic title page (viewed Sep. 16, 2008). "... in partial fulfillment of the requirements for the degree of Masters of Arts in the Department of History." Discipline: History; Department/School: History.
47
KALDRIMIDOU, MARIA. « Images mentales et representations en mathematiques chez des mathematiciens et des etudiants en mathematiques ». Paris 7, 1987. http://www.theses.fr/1987PA077297.
Texte intégral
48
Chen, Chia-Hwan. « Images of the other, images of the self : reciprocal representations of the British and the Chinese from the 1750s to the 1840s ». Thesis, University of Warwick, 2007. http://wrap.warwick.ac.uk/63281/.
Texte intégralRésumé :
During the interactions between the Chinese and the English from the 1750s to the 1840s, writers from both countries have created many distinctive images to represent "the Other" in their own discourses. Imagologists like Jean-Marc Moura (1992) and Daniel-Henri Pageaux (1994) indicated that every image of an "Other" de facto corresponds to an image of "Self." Consequently, the reciprocal images of the British and the Chinese may not only reflect individual writer's attitude towards "the Other" but also refract the self-images of each writer's own people and society. As writers are more or less conditioned by their immediate society, their images of "the Other" tend to reflect the collective ideology of a society. A study of reciprocal images in their own historical milieus will enable one to see why both parties were conditioned to produce certain images to represent "the Other" and why certain images may last longer than the others or even become stereotypes in different discourses. This thesis argues that neither the British nor the Chinese had unanimous images for each other from the 1750s to the 1840s, a century prior to the first Opium War. Instead, writers of both countries had created various negative and positive images of "the Other" to meet their own intentions during this period. By discussing the political, psychological and sociological meanings of the reciprocal images of the British and the Chinese diachronically and synchronically, this thesis suggests that writers might follow certain principles and rules to formulate their own images of other people as "the Other."
49
Bosilj, Petra. « Image indexing and retrieval using component trees ». Thesis, Lorient, 2016. http://www.theses.fr/2016LORIS396/document.
Texte intégralRésumé :
Cette thèse explore l’utilisation de représentations hiérarchiques des images issues de la morphologie mathématique, les arbres des coupes, pour la recherche et la classification d’images. Différents types de structures arborescentes sont analysés et une nouvelle classification en deux superclasses est proposée, ainsi qu’une contribution à l’indexation et à la représentation de ces structures par des dendogrammes. Deux contributions à la recherche d’images sont proposées, l’une sur la détection de régions d’intérêt et l’autre sur la description de ces régions. Les régions MSER peuvent être détectées par un algorithme s’appuyant sur une représentation des images par arbres min et max. L’utilisation d’autres structures arborescentes sous-jacentes permet de détecter des régions présentant des propriétés de stabilité différentes. Un nouveau détecteur, basé sur les arbres des formes, est proposé et évalué en recherche d’images. Pour la description des régions, le concept de spectres de formes 2D permettant de décrire globalement une image est étendu afin de proposer un descripteur local, au pouvoir discriminant plus puissant. Ce nouveau descripteur présente de bonnes propriétés à la fois de compacité et d’invariance à la rotation et à la translation. Une attention particulière a été portée à la préservation de l’invariance à l’échelle. Le descripteur est évalué à la fois en classification d’images et en recherche d’images satellitaires. Enfin, une technique de simplification des arbres de coupes est présentée, qui permet à l’utilisateur de réévaluer les mesures du niveau d’agrégation des régions imposé par les arbres des coupesThis thesis explores component trees, hierarchical structures from Mathematical Morphology, and their application to image retrieval and related tasks. The distinct component trees are analyzed and a novel classification into two superclasses is proposed, as well as a contribution to indexing and representation of the hierarchies using dendrograms. The first contribution to the field of image retrieval is in developing a novel feature detector, built upon the well-established MSER detection. The tree-based implementation of the MSER detector allows for changing the underlying tree in order to produce features of different stability properties. This resulted in the Tree of Shapes based Maximally Stable Region detector, leading to improvements over MSER in retrieval performance. Focusing on feature description, we extend the concept of 2D pattern spectra and adapt their global variant to more powerful, local schemes. Computed on the components of Min/Max-tree, they are histograms holding the information on distribution of image region attributes. The rotation and translation invariance is preserved from the global descriptor, while special attention is given to achieving scale invariance. We report comparable results to SIFT in image classification, as well as outperforming Morphology-based descriptors in satellite image retrieval, with a descriptor shorter than SIFT. Finally, a preprocessing or simplification technique for component trees is also presented, allowing the user to reevaluate the measures of region level of aggregation imposed on a component tree. The thesis is concluded by outlining the future perspectives based on the content of the thesis
50
Garrido, Ostermann Luís. « Hierarchical region based processing of images and video sequences : application to filtering, segmentation and information retrieval ». Doctoral thesis, Universitat Politècnica de Catalunya, 2002. http://hdl.handle.net/10803/6878.
Texte intégralRésumé :
Este trabajo estudia la utilidad de representaciones jerárquicas basadas en regiones para el procesado de imagen y de secuencias de vídeo. Las representaciones basadas en regiones ofrecen una forma de realizar un primer nivel de abstracción y reducir el número de elementos a procesar con respecto a la representación clásica basada en el pixel. En este trabajo se revisan las dos representaciones que han demostrado ser de utilidad para el procesado basado en regiones, a saber el grafo de regiones adyacentes y el árbol, y se discute por qué las representaciones basadas en árboles son más adecuadas para nuestro propósito. De hecho, los árboles permiten la representación de la imagen de forma jerárquica y pueden ser aplicadas sobre éste técnicas eficientes y complejas. En este trabajo se discuten dos cuestiones principales: cómo puede ser creada la representación jerárquica a partir de una imagen determinada y cómo se puede manipular o procesar el árbol.Se han desarrollado dos representaciones basadas en árboles: el Árbol de Máximos, y el Árbol de Particiones Binario. El Árbol de Máximos estructura de forma compacta las componentes conexas que surgen de todos los posibles conjuntos de niveles de una imagen de nivel de gris. Es una representación adecuada para la implementación de operadores conexos antiextensivos, desde operadores clásicos (por ejemplo, filtro de área) hasta operadores nuevos (como el filtro de movimiento desarrollado en este trabajo). El Árbol de Particiones Binario estructura el conjunto de regiones que se obtiene durante la ejecución de un algoritmo de fusión basado en regiones. Desarrollado para superar alguno de los inconvenientes impuestos por el árbol de Máximos -- en particular la falta de flexibilidad de la creación del árbol y la auto-dualidad de la representación del árbol --, ha demostrado ser una representación apta para un gran número de aplicaciones, tal y como se muestra en este trabajo.
Las estrategias de procesado se basan en técnicas de poda. Las técnicas de poda eliminan algunas ramas del árbol basándose en un algoritmo de análisis aplicado a los nodos del árbol. Las técnicas de poda aplicadas al árbol de Máximos permiten obtener operadores anti-extensivos, mientras que para el caso del árbol de Particiones Binario se obtienen operadores auto-duales si éste ha sido creado de forma auto-dual. Las técnicas de poda desarrolladas en este trabajo están dirigidas hacia las siguiente aplicaciones: filtrado, segmentación y recuperación de datos basada en el contenido.
Las aplicaciones de filtrado (en el contexto de los operadores conexos) y segmentación están basados en el mismo principio: los nodos del árbol son analizados de acuerdo a un criterio determinado, y la decisión de eliminar o preservar un nodo se basada normalmente en un umbral aplicado sobre la anterior medida del criterio. La poda se realiza entonces de acuerdo con la ésta decisión. Como resultado, la imagen asociada al árbol podado representa una versión filtrada o segmentada de la imagen original de acuerdo con el criterio seleccionado. Alguno de los criterios discutidos en este trabajo están basados, por ejemplo, en área, movimiento, marcador & propagación o una estrategia de tasa-distorsión. El problema de la falta de robustez de las estrategias clásicas para criterios no crecientes es estudiado y solucionado gracias a un algoritmo de optimización basado en el algoritmo de Viterbi.
La recuperación de imágenes basada en el contenido es la tercera aplicación en la que nos hemos centrado en este trabajo. Las representaciones jerárquicas basadas en regiones son particularmente adecuadas para este propósito ya que permiten representar la imagen a diferentes escalas de resolución, y por lo tanto las regiones asociadas a una imagen pueden ser descritas a diferentes escalas de resolución. En este trabajo nos centramos en un sistema de recuperación de imágenes que soporta preguntas de bajo nivel basadas en descriptores visuales y relaciones espaciales. Para ello, se adjuntan descriptores de región a los nodos del árbol. Se discuten dos tipos de preguntas: pregunta basada en una región, en el que la pregunta esta formada por una región, y pregunta basada en múltiples regiones, en el que la pregunta esta formada por un conjunto de regiones. En el primero la recuperación se realiza utilizando descriptores visuales, mientras que en el segundo se utilizan descriptores visuales y relaciones espaciales. Además, se presenta una estrategia de realimentación por relevancia para eludir la necesidad de establecer manualmente el peso asociado a cada uno de los descriptores.
Un aspecto importante que se ha tenido en cuenta a lo largo de este trabajo es la implementación eficiente de los algoritmos desarrollados tanto para la creación como el procesado del árbol. En el caso de la creación del árbol, la eficiencia se obtiene principalmente gracias al uso de colas jerárquicas, mientras que en el procesado se utilizan algoritmos basados en estrategias recursivas para obtener algoritmos eficientes.
This work discusses the usefulness of hierarchical region based representations for image and video processing. Region based representations offer a way to perform a first level of abstraction and reduce the number of elements to process with respect to the classical pixel based representation. In this work the two representations that have demonstrated to be useful for region based processing are reviewed, namely region adjacency graphs and trees, and it is discussed why tree based representations are better suited for our purpose. In fact, trees allow representing the image in a hierarchical way and efficient and complex processing techniques can be applied on it. Two major issues are discussed in this work: how the hierarchical representation may be created from a given image and how the tree may be manipulated or processed.
Two tree based representations have been developed: the Max-Tree, and the Binary Partition Tree. The Max-Tree structures in a compact way the connected components that arise from all possible level sets from a gray-level image. It is suitable for the implementation of anti-extensive connected operators, ranging from classical ones (for instance, area filter) to new ones (such as the motion filter developed in this work). The Binary Partition Tree structures the set of regions that are obtained during the execution of a region merging algorithm. Developed to overcome some of the drawbacks imposed by the Max-Tree -- in particular the lack of flexibility of the tree creation and the self-duality of the tree representation --, it has demonstrated to be a representation useful for a rather large range of applications, as it is shown in this work.
Processing strategies are focused on pruning techniques. Pruning techniques remove some of the branches of the tree based on an analysis algorithm applied on the nodes of the tree. Pruning techniques applied on the Max-Tree lead to anti-extensive operators, whereas self-dual operators are obtained on the Binary Partition Tree, if the tree is created in a self-dual manner. The pruning techniques that have been developed in this work are directed to the following applications: filtering, segmentation and content based image retrieval.
The filtering (in the context of connected operators) and segmentation applications are based on the same principle: the nodes of the tree are analyzed according to a fixed criterion, and the decision to remove or preserve a node usually relies on a threshold applied on the former measured criterion. Pruning is then performed according to the previous decision. As a result, the image associated to the pruned tree represents a filtered or segmented version of the original image according to the selected criterion. Some of the criteria that are discussed in this work are based, for instance, on area, motion, marker & propagation or a rate-distortion strategy. The problem of the lack of robustness of classical decision approaches of non-increasing criteria is discussed and solved by means of an optimization strategy based on the Viterbi algorithm.
Content based image retrieval is the third application we have focused on in this work. Hierarchical region based representations are particularly well suited for this purpose since they allow to represent the image at different scales of resolution, and thus the regions of the image can be described at different scales of resolution. In this work we focus on an image retrieval system which supports low level queries based on visual descriptors and spatial relationships. For that purpose, region descriptors are attached to the nodes of the tree. Two types of queries are discussed: single region query, in which the query is made up of one region and, multiple region query, in which the query is made up of a set of regions. In the former visual descriptors are used to perform the retrieval whereas visual descriptors and spatial relationships are used in the latter case. Moreover, a relevance feedback approach is presented to avoid the need of manually setting the weights associated to each descriptor.
An important aspect that has been taken into account throughout this work is the efficient implementation of the algorithms that have been developed for both creation and processing of the tree. In the case of the tree creation, efficiency has been obtained mainly due to the use of hierarchical queues, whereas in the processing step analysis algorithms based on recursive strategies are used to get efficient algorithms.