A new bilateral mesh smoothing method by recognizing features

In this paper, we propose a new mesh smoothing method for mesh models with measurement noise obtained from a 3D scanner. Our proposed method enables not only the removal of noise from the faces but the preservation and smoothing of features recognized as sharp edges and corners. First, by using the eigen analysis of normal voting tensor, mesh edges composed of sharp edges can be extracted. Second, bilateral filtering is applied to extracted mesh edges to smooth the mesh edges along the direction of the sharp edge. Finally, by applying modified bilateral filtering to the overall mesh, a mesh model with sharp edges smoothed along their directions and measurement noise removed from faces was obtained.