I have developed and implemented a least-squares theory that makes robust estimates of the angle-dependent reflectivity from recorded surface seismic reflection data. Both angles and reflection coefficients are estimated directly from the data without a priori knowledge of the subsurface structural dip. The implementational cost of this approach is little more than a standard Kirchhoff prestack migration. The method naturally allows compensation for source/receiver directivity, geometric spreading, transmission loss, high-frequency Q attenuation, and limited data acquisition aperture. Synthetic tests show good to excellent results on a 1-D and 2-D model with up to 30 of structural dip. However, amplitude artifacts and errors can be magnified by dipping structure due to increased spatial aliasing of the migration operator, and hyperbolic event interference and coupling at the far offsets.