Repairing 3D Models Obtained From Range Sensors

This paper presents a method to restore 3D models created from range sensors. The approach recovers the 3D information that is lost during the scanning process, usually due to inaccessible zones on the object (occluded areas and crevices). The raw and incomplete 3D data are first projected onto a grid to identify the zones to be repaired. Our algorithm calculates a suitable projection plane for each lost area defined and restores the 2D projected image. Finally, the inverse 2D to 3D transformation is carried out, and the new 3D data are merged with the initial mesh model, providing the completely restored surface. The approach has been tested on a database containing a wide variety of models, yielding excellent results. The experimental section shows that our method works for a large diversity of non-sensed and difficult-to-access zones and provides precise restorations.