Ground-Penetrating Radar (GPR) is widely used on polythermal glaciers to image bed topography and detect internal scatter due to water inclusions in temperate ice. The glaciological importance of this is two-fold: bed topography is a primary component for modelling the long-term evolution of glaciers and ice sheets, and the presence of temperate ice and associated englacial water significantly reduces overall ice viscosity. Englacial water has a direct influence on radar velocity, which can result in incorrect observations of bed topography due to errors in depth conversion. Assessment of radar velocities often requires multi-offset surveys, yet these are logistically challenging and time consuming to acquire, hence techniques to extract velocity from common-offset data are required. We calculate englacial radar velocity from common offset GPR data collected on Von Postbreen, a polythermal glacier in Svalbard. We first separate and enhance the diffracted wavefield by systematically assessing data coherence. We then use the focusing metric of negative entropy to deduce a migration velocity field and produce a velocity model which varies spatially across the glacier. We show that this velocity field successfully differentiates between areas of cold and temperate ice and can detect lateral variations in radar velocity close to the glacier bed. This velocity field results in consistently lower ice depths relative to those derived from a commonly assumed constant velocity, with an average difference of 4.9 ± 2.5% of local ice depth. This indicates that diffraction focusing and velocity estimation are crucial in retrieving correct bed topography in the presence of temperate ice.