Kernel-Based Laplacian Smoothing Method for 3D Mesh Denoising

Hicham Badri (1) , Mohammed El Hassouni (2) , Driss Aboutajdine (3)

1 GeoStat - Geometry and Statistics in acquisition data
2 LRIT
3 GSCM-LRIT - Laboratoire de Recherche en Informatique et Télécommunications [Rabat]

In this paper, we present an improved Laplacian smoothing technique for 3D mesh denoising. This method filters directly the vertices by updating their positions. Laplacian smoothing process is simple to implement and fast, but it tends to produce shrinking and oversmoothing effects. To remedy this problem, firstly, we introduce a kernel function in the Laplacian expression. Then, we propose to use a linear combination of denoised instances. This combination aims to reduce the number of iterations of the desired method by coupling it with a technique that leadsto oversmoothing. Experiments are conducted on synthetic triangular meshes corrupted by Gaussian noise. Results show that we outperform some existing methods in terms of objective and visual quality.

Domaines: Traitement des images [eess.IV]; Vision par ordinateur et reconnaissance de formes [cs.CV]; Synthèse d'image et réalité virtuelle [cs.GR]

Hicham Badri, Mohammed El Hassouni, Driss Aboutajdine. Kernel-Based Laplacian Smoothing Method for 3D Mesh Denoising. 5th International Conference, ICISP 2012, Jun 2012, Agadir, Morocco. pp.77-84. ⟨hal-00763278⟩