Surface quality of medium density fiberboard (MDF), as a very important criterion for further finishing and utilization, was evaluated using a wide range of measuring parameters (instruments, filters, filtering cut-off length, measuring length, measuring resolution, roughness parameters). However, these various approaches make any data comparison rather difficult. Furthermore, filtering the measured data with a simple Gaussian filter has proved unreliable for wood surfaces. A new, more robust approach by using a Robust Gaussian Regression filter together with a previously tested set of measuring and evaluating parameters was used in this paper in order to quantify the surface roughness of MDF manufactured from different species (beech, poplar, birch, Scots pine, mixture 50 % Scots pine and 50 % beech) and using three different grinding disc distances (0.06, 0.15 and 0.6 mm). If manufacturing parameters are kept constant, the processing roughness of MDF is similar, but variable species anatomy and fuzziness effects will give differentiation in total roughness. The results indicated that fibers made of a mixture of wood chips from Scots pine and beech led to the smoothest MDF surface, followed closely by poplar, birch, and Scots pine. The roughest surface was measured for the panels made of beech fibers. Investigating the influence of the grinding disc distance, the roughest MDF surface was obtained when a grinding gap of 0.6 mm was used. However, surface roughness was found to be quite similar applying a grinding disc distance of 0.06 mm slightly lower for 0.15 mm. This finding indicates that a too fine grinding gap does not improve the MDF surface roughness.