The evolution of current manufacturing processes, such as additive manufacturing processes, enables to produce parts with increasingly complex internal and external geometries, to answer functional requirements. This requires an evolution of the measurement methods to analyze the complete part produced. In this context, the use of computed tomography (CT) is increasing. Considering the problem of surface quality control, and also considering the cost of such a measuring system, it becomes necessary to evaluate the capability of tomography techniques to characterize surface geometry despite an inadequate resolution. To this end, the study proposed in this paper aims at assessing the quality of surface roughness characterization by CT in comparison with classical optical measure means. Special attention is given to thresholding, which isnecessary to extract the surface from CT measurements, which are the basis to evaluate roughness parameters. An advanced analysis is also performed to bring out surface typologies that are more appropriate for CT measurements with poor resolution.