Doppler optical coherence tomography (OCT) quantifies axial motion with high precision, whereas lateral motion cannot be detected by a mere evaluation of phase changes. This problem was solved by the introduction of three-beam Doppler OCT, which, however, entails a high experimental effort. Here, we present the numerical analogue to this experimental approach. Phase-stable complex-valued OCT datasets, recorded with full-field swept-source OCT, are filtered in the Fourier domain to limit imaging to different computational subapertures. These are used to calculate all three components of the motion vector with interferometric precision. As known from conventional Doppler OCT for axial motion only, the achievable accuracy exceeds the actual imaging resolution by orders of magnitude in all three dimensions. The feasibility of this method is first demonstrated by quantifying micro-rotation of a scattering sample. Subsequently, a potential application is explored by recording the 3D motion vector field of tissue during laser photocoagulation in ex-vivo porcine retina.