Images obtained with adaptive optics (AO) systems can be improved by using restoration techniques, the AO correction being only partial. However, these methods require an accurate knowledge of the system point spread function (PSF). Adaptive optics systems allow one to estimate the point spread function (PSF) during the science observation. Using data from the wave-front sensor (WFS), a direct estimation of the averaged parallel phase structure and an estimation of the noise on the measurements are provided. The averaged orthogonal phase structure (not seen by the system), and the aliasing covariance are estimated using an end-to-end AO simulation. Finally, the estimated PSF is reconstructed using the algorithm of Veran et al. (1997). 1 However, this reconstruction is non perfect. Several approximations are done (stationary resudual phase, gaussian phase,
simulated aliasing, etc...) and can impact the optical transfer function (OTF) in the case of a rather poor correction. Our aim is to give an error budget for the whole PSF reconstruction process and to link this PSF reconstruction with a deconvolution algorithm that take into account this PSF variability. Indeed, a myopic deconvolution algorithm can be feed with a priori on the object and the PSF. The latter can be obtained by studying the PSF reconstruction error budget as follows in this paper. Finally, this work will lead to an estimation of the error on the deconvolved image allowing one to perform an accurate astrometry/ photometry on the observed objects and to strengthen the contrast in the images. We concluded that to neglect the global cross term or to estimate the aliasing on the measurements using simulations has no effect on the PSF reconstruction.