Nov 16 2017 cs.GR
This paper presents a simple and effective two-stage mesh denoising algorithm, where in the first stage, the face normal filtering is done by using the bilateral normal filtering in the robust statistics framework. \textitTukey's bi-weight function is used as similarity function in the bilateral weighting, which is a robust estimator and stops the diffusion at sharp edges which helps to retain features and removes noise from flat regions effectively. In the second stage, an edge-weighted Laplace operator is introduced to compute the differential coordinate. This differential coordinate helps the algorithm to produce a high-quality mesh without any face normal flips and also makes the method robust against high-intensity noise.