Feature preserving smoothing method for noised meshes

This paper presents a smoothing method for a triangular mesh with feature preservation. In order to obtain a smooth surface from a scanned mesh where noise is contained because of measurement errors, we developed a feature preserving smoothing method. At first, we simplified the scale-dependent discrete Laplacian and defined boundary Laplacian to deal with an open mesh. Then, a method for feature detection is constructed. Furthermore, utilizing this feature detection method to determine a suitable parameter, we have developed a feature preserving smoothing method based on anisotropic diffusion. Vertices that consist the feature lines are detected by our feature detection method, and smoothing is conducted for the whole mesh including feature lines. We demonstrate the presented method and the results give a well-smoothed mesh with its features preserved.