Since its invention in the early seventies, the Shack-Hartmann wavefront sensor has seen a wide variety of applications and has had great success in the fields of Adaptive Optics and Ophthalmology, where interferometry is usually impractical. Its application to optical shop testing has been less visible perhaps because shop environments can be manipulated to sufficiently remove vibration and turbulence to a degree that can support interferometry. However, with the growing need to accurately test aspheric optics, the Shack-Hartmann has an advantage; its dynamic range can be manipulated through the design of the lenslet array, rather than being directly tied to the wavelength of light and therefore lessen the need for expensive null optics.When the Shack-Hartmann is pushed to the limits of dynamic range, several issues must be dealt with. First, to reach the limits of dynamic range, those limits must be well understood. This dissertation presents a graphical approach to designing the Shack-Hartmann sensor that makes the trade-off between sensitivity and dynamic range, and accuracy and resolution intuitively clear. Next, the spots that once landed neatly in the region behind each lenslet, may now wander several lenslets away and the data reduction must be able handle this. This dissertation presents a novel and robust method for sorting these widely wondering spots and is shown to work in measurements of highly aspheric elements. Finally, in the high dynamic range regime, induced aberrations can severely limit the accuracy of the instrument. In this dissertation, these non-linear and measurement-dependent errors are studied in detail and a method of compensation is presented along with experimental results that illustrate the efficacy of the approach.

L'Europe prépare actuellement le plus grand télescope du monde : l'European Extremely Large Telescope (E-ELT). Prévu vers 2026, ce télescope géant permettra de répondre à des questions fondamentales de l'astrophysique contemporaine. L'imagerie d'objets astrophysiques depuis des télescopes au sol est cependant perturbée par l'atmosphère qui réduit la capacité des instruments au sol à distinguer les objets proches. L'Optique Adaptative (OA) permet de restaurer cette résolution angulaire en corrigeant en temps réel (via un miroir déformable) le front d'onde perturbé par l'atmosphère (mesuré par l'Analyseur de Surface d'Onde (ASO)). Jusqu'à récemment, la majorité des systèmes d'OA utilisaient des ASO Shack-Hartmann (SH). Des concepts concurrents basés sur le filtrage optique de Fourier (le senseur Pyramide ou l'analyseur Zernike) viennent cependant d'être mis en fonctionnement et leurs résultats semblent surpasser les performances du SH. En vue de leur potentielle utilisation sur les ELTs, cette thèse vise à consolider leur compréhension théorique ainsi qu'à optimiser ces ASO basés sur le filtrage de Fourier. Cette thèse développe un cadre mathématique qui décrit sous un unique formalisme ces ASO. Il permet de généraliser les designs préexistants -passant ainsi d'ASO uniques à des "classes d'ASO"- en transformant leurs grandeurs caractéristiques à l'origine fixées en degrés de liberté. Les classes Pyramide et Zernike sont donc explorées dans le but d'optimiser ces ASO au regard des attentes expérimentales. Des configurations inédites de la classe Pyramide -ASO que l'on appelle Pyramides aplaties- s'avèrent notamment prometteuses et font l'objet d'une étude plus poussée. / Europe is currently preparing the largest telescope of the world: the European Extremely Large Telescope (E-ELT). Planned by 2026, this huge telescope will allow to answer fundamental questions of contemporary astrophysics. However, images of astrophysical objects done by ground based telescopes suffer from the atmospheric turbulence which reduces the capacity of instruments to distinguish objects too close to each other. The Adaptive Optics (AO) allows to restore this loss of angular resolution by correcting (thanks to a deformable mirror) in real time the perturbed wave front (measured by the WaveFront Sensor (WFS)).Until very recently, the majority of AO systems had used the Shack-Hartmann (SH) WFS. New concepts based on Fourier filtering (the Pyramid or the Zernike WFSs) have however just been put in operation in several professional observatories and their results seem to outperform the SH. Since they would potentially be chosen for the AO systems of the future ELTs, this thesis aims to consolidate their theoretical understanding and to optimize these Fourier based WFSs.We firstly develop a mathematical framework which describes all these WFSs under an unique formalism. It allows to generalize the pre-existent designs -a WFS thus becomes a "WFS class"- by considering their optical parameters as flexible quantities. We then explored the two Pyramid and Zernike classes to identify the influence of class' parameters on performance criteria in order to optimize optical designs with regard to the instrumental requirements. New configurations of the Pyramid class -that we called Flattened pyramids- show promising behaviors and are studied in details.