The point spread function (PSF) of an adaptive optics (AO) system evolves in larger region than isoplanatic patch in the field of view (FOV). This variation strongly limits the efficient utilization of AO images. A theoretical expression of this space-varying PSF is derived by analysis of characteristics of AO imaging system. Then by combining the space-field statistical characteristics of AO compensation and atmospheric turbulence theory a predicted space-varying parameter PSF model for AO systems is developed and the effectiveness of the model is also tested by employing the predicted space-varying PSF model in reconstruction of AO astronomical image with deconvolution method, which can reduce effects of residual wave-front distortion and anisoplanatism on AO system′s partial compensated astronomical images.