When several different velocity models are options for a given dataset, they can be quickly tested using wavefields synthesized from an initial image. Incorporating prestack velocity information from the initial image into both an areal source function and a Born-modeled receiver wavefield allows errors in the initial model to be taken into account and rectified when a more accurate model is tested. Finally, by imaging only isolated locations along a reflector, tests can be conducted in a fraction of the time required for a full migration by migrating only a single shot. Tests on data from the Sigsbee synthetic model demonstrated that this method can effectively discriminate between models, even if they differ only slightly in terms of salt shape or velocity, or if the initial image was created with an inaccurate model. Finally, a quantitative measure of image quality helped identify the most accurate model when qualitative comparisons were inconclusive.