Image super-resolution technique mainly aims at restoring high-resolution image with satisfactory novel details. In recent years, leaning-based single-image super-resolution has been developed and proved to produce satisfactory results. With one or some dictionaries trained from a training set, learning-based super-resolution is able to establish a mapping relationship between low-resolution images and their corresponding high-resolution ones. Among all these algorithms, sparsity-based super-resolution has been proved with outstanding performance from extensive experiments. By utilizing compact dictionaries, this class of super-resolution algorithms can be efficient with lower computation complexity and has shown great potential for the practical applications. Our proposed model, which is known as Joint Dictionary-based Super-Resolution (JDSR) algorithm, is a new sparsity-based super-resolution approach. Based on the observation that the initial values of Non-locally Centralized Sparse Representation (NCSR) model will affect the final reconstruction, we change its initial values by using results of Zeyde's model. Besides, with the purpose of further improvement, we also add a gradient histogram preservation term in the sparse model of NCSR, and modify the reference histogram estimation by a simple edge detection based enhancement so that the estimated histogram will be closer to the ground truth. The experimental results illustrate that our method outperforms the state-of-the-art methods in terms of sharper edges, clearer textures and better novel details.

Cette thèse porte sur la localisation de sources spatialement distribuées. Premièrement, des résultats théoriques ont été établis concernant les performances des méthodes à haute résolution en présence d'erreurs de modèle dues à la distribution angulaire de source. Ainsi, le biais d'estimation et l'erreur quadratique moyenne sont exprimées en fonction des paramètres liés à l'erreur de modèle. De plus, les performances ont été étudiées en fonction de la géométrie de l'antenne afin de déterminer les configurations les plus robustes aux sources dispersées.Les résultats théoriques ont été validés par des simulations numériques. Dans un deuxième temps, une nouvelle approche non paramétrique a été proposée pour la localisation de sources distribuées : cette approche exploite les propriétés de parcimonie et de rang-faible de la matrice de covariance spatiale des sources. Cette méthode permet en outre d'estimer la forme de la distribution spatiale des sources. Les résultats de simulations ont permis de mettre en avant l'intérêt de l'hypothèse rang faible par rapport aux critères exploitant uniquement la parcimonie / This thesis focuses on the distributed source localization problem. In a first step, performance of high resolution methods in the presence of model errors due to the angular distribution of source has been studied. Theoretical expressions of the estimation bias and the mean square error of the direction of arrival of sources have been established in terms of model error. The impacts of the array geometry on the performances have studied in order to optimize the robustness of the array to the model error due to distributed sources.Theoretical results have been validated by numerical simulations.In a second step, a new approach for the localization of spatially distributed source has been proposed: the approach is based on the sparsity and low-rank property of the spatial covariance matrix of the sources. The proposed method provides also an estimation of the distribution shapes of the sources. Simulation results exhibit the advantages of exploiting the sparsity and the low rank properties.