Image inversion is a technique whereby computer-generated images based on an optical system model are compared with images from the real system in a nonlinear least squares estimation procedure. The differences are used to change model parameters until the model-based images match the actual images. Image inversion can be used with a simple exit-pupil optical model, using a single Fourier transform to generate point-source images. This “phase retrieval” approach is effective for determining overall wavefront aberrations, but not for directly determining individual optical element alignments or surface aberrations.