The main objective of this thesis is the automatic modeling, understanding and segmentation of diffusively infiltrative tumors known as Diffuse Low-Grade Gliomas. Two approaches exploiting anatomical and spatial prior knowledge have been proposed. We first present the construction of a tumor specific probabilistic atlas describing the tumors' preferential locations in the brain. The proposed atlas constitutes an excellent tool for the study of the mechanisms behind the genesis of the tumors and provides strong spatial cues on where they are expected to appear. The latter characteristic is exploited in a Markov Random Field based segmentation method where the atlas guides the segmentation process as well as characterizes the tumor's preferential location. Second, we introduce a concurrent tumor segmentation and registration with missing correspondences method. The anatomical knowledge introduced by the registration process increases the segmentation quality, while progressively acknowledging the presence of the tumor ensures that the registration is not violated by the missing correspondences without the introduction of a bias. The method is designed as a hierarchical grid-based Markov Random Field model where the segmentation and registration parameters are estimated simultaneously on the grid's control point. The last contribution of this thesis is an uncertainty-driven adaptive sampling approach for such grid-based models in order to ensure precision and accuracy while maintaining robustness and computational efficiency. The potentials of both methods have been demonstrated on a large data-set of heterogeneous Diffuse Low-Grade Gliomas. The proposed methods go beyond the scope of the presented clinical context due to their strong modularity and could easily be adapted to other clinical or computer vision problems.

The underlying aim of this research is to investigate the mathematical descriptions of homogeneous textures in digital images for the purpose of segmentation and recognition. The research covers the problem of testing these mathematical descriptions by using them to generate synthetic realisations of the homogeneous texture for subjective and analytical comparisons with the source texture from which they were derived. The application of this research is in analysing satellite or airborne images of the Earth's surface. In particular, Synthetic Aperture Radar (SAR) images often exhibit regions of homogeneous texture, which if segmented, could facilitate terrain classification. In this thesis we present noncausal, nonparametric, multiscale, Markov random field (MRF) models for recognising and synthesising texture. The models have the ability to capture the characteristics of, and to synthesise, a wide variety of textures, varying from the highly structured to the stochastic. For texture synthesis, we introduce our own novel multiscale approach incorporating a new concept of local annealing. This allows us to use large neighbourhood systems to model complex natural textures with high order statistical characteristics. The new multiscale texture synthesis algorithm also produces synthetic textures with few, if any, phase discontinuities. The power of our modelling technique is evident in that only a small source image is required to synthesise representative examples of the source texture, even when the texture contains long-range characteristics. We also show how the high-dimensional model of the texture may be modelled with lower dimensional statistics without compromising the integrity of the representation. We then show how these models -- which are able to capture most of the unique characteristics of a texture -- can be for the ``open-ended'' problem of recognising textures embedded in a scene containing previously unseen textures. Whilst this technique was developed for the practical application of recognising different terrain types from Synthetic Aperture Radar (SAR) images, it has applications in other image processing tasks requiring texture recognition.

Thesis (Ph. D.)--Ohio State University, 2008. / Title from first page of PDF file. Includes bibliographical references (p. 140-150).

Dissertation (Ph.D.)--Auburn University, 2007. / Abstract. Vita. Includes bibliographic references (p.111-117).

Thesis advisor: Cowles, Mary K. Includes bibliographic references (p. 167-172).

No description available.

In this thesis, we aim to investigate the correlation between myelination and the gestational age for preterm infants, with the former being an important developmental process during human brain maturation. Quantification of myelin requires dedicated imaging, but the conventional magnetic resonance images routinely acquired during clinical imaging of neonates carry signatures that are thought to be associated with myelination. This thesis thus focuses on structural segmentation and spatio-temporal modelling of the so-called myelin-like signals on T2-weighted scans for early prognostic evaluation of the preterm brain. The segmentation part poses the major challenges of this task: insufficient spatial prior information of myelination and the presence of substantial partial volume voxels in clinical data. Specific spatial priors for the developing brain are obtained from either probabilistic atlases or manually annotated training images, but none of them currently include myelin as an individual tissue type. This causes further difficulties in partial volume estimation which depends on the probabilistic atlases of the composing pure tissues. Our key contribution is the development of an expectation-maximisation framework that incorporates an explicit partial volume class whose locations are configured in relation to the composing pure tissues in a predefined region of interest via second-order Markov random fields. This approach resolves the above challenges without requiring any probabilistic atlas of myelin. We also investigate atlas-based whole brain segmentation that generates the binary mask for the region of interest. We then construct a spatio-temporal growth model for myelin-like signals using logistic regression based on the automatic segmentations of 114 preterm infants aged between 29 and 44 gestational weeks. Lastly, we demonstrate the ability of age estimation using the normal growth model in a leave-one-out procedure.

Nous présentons dans cette thèse un nouveau modèle statistique de forme et l'utilisons pour la segmentation d'images avec a priori. Ce modèle est représenté par un champ de Markov. Les noeuds du graphe correspondent aux points de contrôle situés sur le contour de la forme géométrique, et les arêtes du graphe représentent les dépendances entre les points de contrôle. La structure du champ de Markov est déterminée à partir d'un ensemble de formes, en utilisant des techniques d'apprentissage de variétés et de groupement non-supervisé. Les contraintes entre les points sont assurées par l'estimation des fonctions de densité de probabilité des longueurs de cordes normalisées. Dans une deuxième étape, nous construisons un algorithme de segmentation qui intègre le modèle statistique de forme, et qui le relie à l'image grâce à un terme région, à travers l'utilisation de diagrammes de Voronoi. Dans cette approche, un contour de forme déformable évolue vers l'objet à segmenter. Nous formulons aussi un algorithme de segmentation basé sur des détecteurs de points d'intérêt, où le terme de régularisation est lié à l'apriori de forme. Dans ce cas, on cherche à faire correspondre le modèle aux meilleurs points candidats extraits de l'image par le détecteur. L'optimisation pour les deux algorithmes est faite en utilisant des méthodes récentes et efficaces. Nous validons notre approche à travers plusieurs jeux de données en 2D et en 3D, pour des applications de vision par ordinateur ainsi que l'analyse d'images médicales. / In this thesis, we introduce a new statistical shape model and use it for knowledge-based image segmentation. The model is represented by a Markov Random Field (MRF). The vertices of the graph correspond to landmarks lying on the shape boundary, whereas the edges of the graph encode the dependencies between the landmarks. The MRF structure is determined from a training set of shapes using manifold learning and unsupervised clustering techniques. The inter-point constraints are enforced using the learnedprobability distribution function of the normalized chord lengths.This model is used as a basis for knowledge-based segmentation. We adopt two approaches to incorporate the data support: one is based on landmark correspondences and the other one uses image region information. In the first case, correspondences between the model and the image are obtained through detectors and the optimal configuration is achieved through combination of detector responses and prior knowledge. The second approach consists of minimizing an energy that discriminates the object from the background while accounting for the shape prior. A Voronoi decomposition is used to express this objective function in a distributed manner using the landmarks of the model. Both algorithms are optimized using state-of-the art eficient optimization methods. We validate our approach on various 2D and 3D datasets of images, for computer vision applications as well as medical image analysis.

Cette thèse est dédiée au développement de méthodes à base de graphes, permettant de traiter les problèmes fondamentaux de la vision par ordinateur tels que la segmentation, le suivi d’objets, l’appariement de formes et l’inférence de modèles 3D. La première contribution de cette thèse est une méthode unifiée reposant sur un champ de Markov aléatoire (MRF) d’ordre deux permettant de réaliser en une seule étape la segmentation et le suivi de plusieurs objets observés par une caméra unique, tout en les ordonnançant en fonction de leur distance à la caméra. Nous y parvenons au moyen d’un nouveau modèle stratifié (2.5D) dans lequel une représentation bas-niveau et une représentation haut-niveau sont combinées par le biais de contraintes locales. Afin d’introduire des connaissances de haut niveau a priori, telles que des a priori sur la forme des objets, nous étudions l’appariement non-rigide de surfaces 3D. La seconde contribution de cette thèse consiste en une formulation générique d’appariement de graphes qui met en jeu des potentiels d’ordre supérieur et qui est capable d’intégrer différentes mesures de similarités d’apparence, de similarités géométriques et des pénalisations sur les déformations des formes. En tant que la troisième contribution de cette thèse, nous considérons également des interactions d’ordre supérieur pour proposer un a priori de forme invariant par rapport à la pose des objets, et l’exploitons dans le cadre d’une nouvelle approche de segmentation d’images médicales 3D afin d’obtenir une méthode indépendante de la pose de l’objet d’intérêt et de l’initialisation du modèle de forme. La dernière contribution de cette thèse vise à surmonter l’influence de la pose de la caméra dans les problèmes de vision. Nous introduisons un paradigme unifié permettant d’inférer des modèles 3D à partir d’images 2D monoculaires. Ce paradigme détermine simultanément le modèle 3D optimal et les projections 2D correspondantes sans estimer explicitement le point de vue de la caméra, tout en gérant les mauvaises détections et les occlusions. / This thesis is devoted to the development of graph-based methods that address several of the most fundamental computer vision problems, such as segmentation, tracking, shape matching and 3D model inference. The first contribution of this thesis is a unified, single-shot optimization framework for simultaneous segmentation, depth ordering and multi-object tracking from monocular video sequences using a pairwise Markov Random Field (MRF). This is achieved through a novel 2.5D layered model where object-level and pixel-level representations are seamlessly combined through local constraints. Towards introducing high-level knowledge, such as shape priors, we then studied the problem of non-rigid 3D surface matching. The second contribution of this thesis consists of a higher-order graph matching formulation that encodes various measurements of geometric/appearance similarities and intrinsic deformation errors. As the third contribution of this thesis, higher-order interactions were further considered to build pose-invariant statistical shape priors and were exploited for the development of a novel approach for knowledge-based 3D segmentation in medical imaging which is invariant to the global pose and the initialization of the shape model. The last contribution of this thesis aimed to partially address the influence of camera pose in visual perception. To this end, we introduced a unified paradigm for 3D landmark model inference from monocular 2D images to simultaneously determine both the optimal 3D model and the corresponding 2D projections without explicit estimation of the camera viewpoint, which is also able to deal with misdetections/occlusions

La présente thèse est consacrée au recalage et à la fusion d’images de façon dense et déformable via des méthodes d’optimisation discrète. La contribution majeure consiste en un principe de couplage entre recalage géométrique et iconique via l’utilisation de méthodes dites graphiques. Une telle formulation peut être obtenue à partir d’un Champ de Markov Aléatoire binaire et permet de résoudre les deux problèmes simultanément tout en imposant une cohérence à leurs solutions respectives. La méthodologie s’applique à la fusion de paires d’images (dans ses versions symétrique et asymétrique), ainsi qu’au recalage simultané de groupes d’images nécessaire à l’étude de populations. Les qualités principales de notre approche résident dans sa faible complexité algorithmique et sa versatilité. L’utilisation d’une formulation discrète assure une grande modularité concernant tant la mesure de similarité iconique que l’extraction et l’association de points d’intérêt. Les résultats prometteurs obtenus sur les bases de données de référence en flot optique et sur des données médicales tridimensionnelles démontrent tout le potentiel de notre méthodologie / This thesis is devoted to dense deformable image registration/fusion using discrete methods. The main contribution of the thesis is a principled registration framework coupling iconic/geometric information through graph-based techniques. Such a formulation is derived from a pair-wise MRF view-point and solves both problems simultaneously while imposing consistency on their respective solutions. The proposed framework was used to cope with pair-wise image fusion (symmetric and asymmetric variants are proposed) as well as group-wise registration for population modeling. The main qualities of our framework lie in its computational efficiency and versatility. The discrete nature of the formulation renders the framework modular in terms of iconic similarity measures as well as landmark extraction and association techniques. Promising results using a standard benchmark database in optical flow estimation and 3D medical data demonstrate the potentials of our methods.