As new functional requirements of products lead to the definition of more complicated shapes, reverse engineering is playing a more important role. The process consists in defining a CAD model of the object surfaces from the measurement of the real object. Reverse engineering takes advantage of new advances in noncontact measuring systems leading to a representation of the surfaces as large clouds of points. Nevertheless, scanning without path planning may affect completeness and accuracy of the measured data. This paper addresses the problem of intelligent scan planning within the context of reverse engineering. A measuring system allows us to acquire a cloud of points, which represents the first measurement of the free-form object. This incomplete and locally inaccurate cloud of points is used as a basis to generate an intelligent scan planning. A pretreatment of the point cloud is performed to determine the quality of the first scan and to find out the characteristic edges. The method relies on a voxel representation of the data. According to given thresholds of quality criteria (noise and completeness), unsatisfactory quality zones and digitizing gaps are identified. The new scan paths for an optimal digitizing are then calculated including optimal orientation search. An experimental application of the presented work is described through the digitizing of a face mask.