INTRODUCTION – Magnetic resonance phase data directly reflects local magnetic field changes and is used for shimming procedures, interpretation or correction of artifacts, and as anatomical contrast complementary to magnitude images [1]. Most of these applications require separation of local phase and background phase contributions that result from air-tissue and bone-tissue interfaces, eddy currents, temperature fluctuations or gradient deviations. Although several heuristic preprocessing methods [1, 2] have been established for this non-trivial task, it has been shown that these methods exhibit significant concomitant degradation of local phase information [3]. In this contribution, we present, for the first time, a non-heuristic, parameter-free approach for high-precision separation of local phase and background phase contributions for in vivo SWI-data.