Non-Local Low-Rank Normal Filtering for Mesh Denoising

by Xianzhi Li, Lei Zhu, Chi-Wing Fu and Pheng-Ann Heng.

Introduction

This repository is for our Pacific Graphics 2018 paper 'Non-Local Low-Rank Normal Filtering for Mesh Denoising'. In this paper, we present a non-local low-rank normal filtering method for mesh denoising. By exploring the geometric similarity between local surface patches on 3D meshes in the form of normal fields, we devise a low-rank recovery model that filters normal vectors by means of patch groups.

In this repository, we release demo (exe files), code (C++), and data.

Citation

If you find our work useful in your research, please consider citing:

@article{xianzhi2018nllr, 
 title={Non-local low-rank normal filtering for mesh denoising}, 
 author={Li, Xianzhi and Zhu, Lei and Fu, Chi-Wing and Heng, Pheng-Ann},
 journal={Computer Graphics Forum (Pacific Graphics)}, 
 volume={37},
 number={7},
 pages={155--166},
 year={2018}
}

Usage

To try our method for mesh denoising, you can directly run the 'NLLR.exe' inside demo.rar.

upzip the demo.rar
copy noisy mesh into the demo folder (e.g., child_n3.off)
run: NLLR.exe child_n3.off 0.39 10 10

'child_n3.off' is the input noisy mesh, 0.39 is the sigma_M in our paper, 10 is the N_k and 10 is the number of vertex updating. You will see the denoised mesh inside the same folder. For the 'NLLR.exe', you can refer to NLLR folder for the source code.

To evaluate the denoising performance, you can directly run the 'evaluation.exe' inside demo.rar.

upzip the demo.rar
copy ground truth mesh into the demo folder (e.g., child.off)
run: evaluation.exe child.off denoised_result.off

'child.off' is the ground truth mesh, 'denoised_result.off' is the denoised mesh. You will see the mean square angle error (MSAE).

Questions

Please contact '[email protected]'