Abstract : The performance of an adaptive optics (AO) system is intrinsically related to its capacity to measure the deformations of the incoming wavefront its propagation trough the atmosphere. However some limitations restrict the efficiency of this wavefront analysis. The main ones are the measurement noise and the anisoplanatism. The first one limits the magnitude of the astronomical objects that are likely to guide the AO systems, while the second one restricts their corrected field to a few tens of arcseconds, due to the volumic distribution of the atmospheric turbulence. Together, these limitations reduce to a few percents the sky coverage of the instruments assisted by AO. For this reason, rising generation of AO systems aims at either the very high performance, implying in particular a good robustness to measurement noise, or the widening of the corrected field, implying the knowledge of the 3D distribution of the atmospheric turbulence. For these wide field AO systems, it is essential to use several directions of analysis. The wavefront analysis problem is then articulated around three items: the guides stars one has to carry out for the multi-analysis, the wavefront sensing concept considered to measure the volume of turbulence and the capacity of the wavefront sensors to provide good measurements. The purpose of this PhD thesis is to study each one of these three aspects. Hence a redefinition of the concept of sky coverage is proposed for wide field AO. The definition suggested here takes into account the number of guide stars and their magnitude, but also the fraction of the scientific field that is really covered, the considered wavefront sensing concept and the expected performance of the instrument. Concerning the wavefront sensing concepts, a comparative study of the Star Oriented and Layer Oriented concepts is presented and an optimization is proposed for each one, leading to very close performance. Finally, a comparison of different wavefront slope estimators is proposed for the Shack-Hartmann wavefront sensor. In particular, a weighted center of gravity estimator is studied, that offers a good robustness to noise and good properties of linearity. This last study is usefull as well for wide field AO as for very high performance AO.