Yadav SK, Reitebuch U, Polthier K. "Robust and high fidelity mesh denoising." IEEE Trans Vis Comput Gr 2018. 1–1. doi: 10.1109/TVCG.2018.2828818 .
Abstract: This paper presents a simple and effective two-stage mesh denoising algorithm, where in the first stage, face normal filtering is done by using bilateral normal filtering in a robust statistics framework. Tukey’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 to retain features and removes noise from flat regions effectively. In the second stage, an edge-weighted Laplace operator is introduced to compute a differential coordinate. This differential coordinate helps the algorithm to produce a high-quality mesh without any face normal flips and makes the method robust against high-intensity noise.
Windows 7 or newer.
Java.
obj, off, stl and jvx.
To compile the provided sourec code:
-
Please extract the folder RoFi.zip.
-
Double click on launcher.bat file.
-
It will open a new window without any model.
-
Click on the "Browse Computer" button and upload a nosiy mesh (.stl, .off, .obj and .jvx).
-
Folder has a noisy fandisk model (user can load that model too) .
-
There are three different parameters to tune: range variation (\sigma_s), isotropic factor (\lambda_I) and number of iterations (p).
-
Tune the parameters according to the amount of noise and desired output.
-
"Reset" button will restore the original noisy model.
-
"Mesh quality" button computes the mesh quality factor Q of the current geometry.
If you have further questions or not able to compile the algorithm, please write me on following emails: