Joint Bilateral Mesh Denoising using Color Information and Local Anti-Shrinking

Oliver Wasenmüller, Gabriele Bleser, Didier Stricker

In: Vaclav Skala (editor). Journal of WSCG Proceedings of the International Conference on Computer Graphics, Visualization and Computer Vision. 23 Pages 27-34 6/2015.


Recent 3D reconstruction algorithms are able to generate colored meshes with high resolution details of given objects. However, due to several reasons the reconstructions still contain some noise. In this paper we propose the new Joint Bilateral Mesh Denoising (JBMD), which is an anisotropic filter for highly precise and smooth mesh denoising. Compared to state of the art algorithms it uses color information as an additional constraint for denoising; following the observation that geometry and color changes often coincide. We face the well-known mesh shrinking problem by a new local anti-shrinking, leading to precise edge preservation. In addition we use an increasing smoothing sensitivity for higher numbers of iterations. We show in our evaluation with three different categories of testdata that our contributions lead to high precision results, which outperform competing algorithms. Furthermore, our JBMD algorithm converges on a minimal error level for higher numbers of iterations.


JBMD.pdf (pdf, 21 MB )

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz