The Mean Dynamic Topography (MDT) of the ocean provides valuable information about the ocean's surface currents. Therefore the MDT is computed from satellite observations and then assimilated into ocean models in order to improve the ocean circulation estimates. However, the computation of the MDT from satellite observations of sea surface height and the Earth's gravity field is not straightforward and requires additional filtering of the data combination. The choice of the filter is crucial as it determines the amount of small-scale noise in the data and the resolution of the final MDT. There exist various approaches for the determination of an “optimal” filter. However, they all have in common the more or less subjective choice of the filter type and filter width. Here, a new filter is presented that is determined directly from the geodetic normal equations. By its construction, this filter accurately accounts for the correlations within the MDT data and requires no subjective choice about the filter radius. The new filtered MDT is assimilated into an inverse ocean model. Modifications in the meridional overturning circulation and in the poleward heat transports can be observed, compared to the result of the assimilation using the unfiltered MDT.