Deep learning has overwhelmingly impacted post-acquisition image-processing tasks, however, there is increasing interest in more tightly coupled computational imaging approaches, where models, computation, and physical sensing are intertwined. This dissertation focuses on how to leverage the expressive power of deep learning in image reconstruction. We use deep learning in both the sensor data domain and the image domain to develop new fast and efficient algorithms to achieve superior quality imagery. Metal artifacts are ubiquitous in both security and medical applications. They can greatly limit subsequent object delineation and information extraction from the images, restricting their diagnostic value. This problem is particularly acute in the security domain, where there is great heterogeneity in the objects that can appear in a scene, highly accurate decisions must be made quickly, and the processing time is highly constrained. Motivated primarily by security applications, we present a new deep-learning-based MAR approach that tackles the problem in the sensor data domain. We treat the observed data corresponding to dense, metal objects as missing data and train an adversarial deep network to complete the missing data directly in the projection domain. The subsequent complete projection data is then used with an efficient conventional image reconstruction algorithm to reconstruct an image intended to be free of artifacts. Conventional image reconstruction algorithms assume that high-quality data is present on a dense and regular grid. Using conventional methods when these requirements are not met produces images filled with artifacts that are difficult to interpret. In this context, we develop data-domain deep learning methods that attempt to enhance the observed data to better meet the assumptions underlying the fast conventional analytical reconstruction methods. By focusing learning in the data domain in this way and coupling the result with existing conventional reconstruction methods, high-quality imaging can be achieved in a fast and efficient manner. We demonstrate results on four different problems: i) low-dose CT, ii) sparse-view CT, iii) limited-angle CT, and iv) accelerated MRI. Image domain prior models have been shown to improve the quality of reconstructed images, especially when data are limited. A novel principled approach is presented allowing the unified integration of both data and image domain priors for improved image reconstruction. The consensus equilibrium framework is extended to integrate physical sensor models, data models, and image models. In order to achieve this integration, the conventional image variables used in consensus equilibrium are augmented with variables representing data domain quantities. The overall result produces combined estimates of both the data and the reconstructed image that is consistent with the physical models and prior models being utilized. The prior models used in both image and data domains in this work are created using deep neural networks. The superior quality allowed by incorporating both data and image domain prior models is demonstrated for two applications: limited-angle CT and accelerated MRI. A major question that arises in the use of neural networks and in particular deep networks is their stability. That is, if the examples seen in the application environment differ from the training environment will the performance be robust. We perform an empirical stability analysis of data and image domain deep learning methods developed for limited-angle CT reconstruction. We consider three types of perturbations to test stability: adversarially optimized, random, and structural perturbations. Our empirical analysis reveals that the data-domain learning approach proposed in this dissertation is less susceptible to perturbations as compared to the image-domain post-processing approach. This is a very encouraging result and strongly supports the main argument of this dissertation that there is value in using data-domain learning and it should be a part of our computational imaging toolkit.

Les expériences sismiques actives sont largement utilisées pour caractériser la structure de la subsurface. Les méthodes dites d’analyse de vitesse par migration ont pour but la détermination d’un macro-modèle de vitesse, lisse, et contrôlant la cinématique de propagation des ondes. Le modèle est estimé par des critères de cohérence d’image ou de focalisation d’image. Les images de réflectivité obtenues par les techniques de migration classiques sont cependant contaminées par des artefacts, altérant la qualité de la remise à jour du macro-modèle. Des résultats récents proposent de coupler l’inversion asymptotique, qui donne des images beaucoup plus propres en pratique, avec l’analyse de vitesse pour la version offset en profondeur. Cette approche cependant demande des capacités de calcul et de mémoire importantes et ne peut actuellement être étendue en 3D.Dans ce travail, je propose de développer le couplage entre l’analyse de vitesse et la migration plus conventionnelle par point de tir. La nouvelle approche permet de prendre en compte des modèles de vitesse complexes, comme par exemple en présence d’anomalies de vitesses plus lentes ou de réflectivités discontinues. C’est une alternative avantageuse en termes d’implémentation et de coût numérique par rapport à la version profondeur. Je propose aussi d’étendre l’analyse de vitesse par inversion au domaine des données pour les cas par point de tir. J’établis un lien entre les méthodes formulées dans les domaines données et images. Les méthodologies sont développées et analysées sur des données synthétiques 2D. / Active seismic experiments are widely used to characterize the structure of the subsurface. Migration Velocity Analysis techniques aim at recovering the background velocity model controlling the kinematics of wave propagation. The first step consists of obtaining the reflectivity images by migrating observed data in a given macro velocity model. The estimated model is then updated, assessing the quality of the background velocity model through the image coherency or focusing criteria. Classical migration techniques, however, do not provide a sufficiently accurate reflectivity image, leading to incorrect velocity updates. Recent investigations propose to couple the asymptotic inversion, which can remove migration artifacts in practice, to velocity analysis in the subsurface-offset domain for better robustness. This approach requires large memory and cannot be currently extended to 3D. In this thesis, I propose to transpose the strategy to the more conventional common-shot migration based velocity analysis. I analyze how the approach can deal with complex models, in particular with the presence of low velocity anomaly zones or discontinuous reflectivities. Additionally, it requires less memory than its counterpart in the subsurface-offset domain. I also propose to extend Inversion Velocity Analysis to the data-domain, leading to a more linearized inverse problem than classic waveform inversion. I establish formal links between data-fitting principle and image coherency criteria by comparing the new approach to other reflection-based waveform inversion techniques. The methodologies are developed and analyzed on 2D synthetic data sets.

This thesis deals with the domain indexes in Oracle Database 11g. It describes the database architecture and discusses the available methods of indexing. There are explained concrete ways of the implementation and use of domain indexes, also discussed ways of indexing spatio-temporal data especially the TB-tree structure, which is then implemented as a domain index. Along with the domain index operators are also implemented by means of which the index is subsequently used and tested.