Rolling normal filtering for point clouds

3D geometric features represent rich details of 3D models, whose scale is much larger than noise. Traditional point cloud denoising methods cannot handle the task of processing and analyzing these features. Rolling guidance normal filtering is proved to be a useful tool in image and mesh small features removing. However, its direct extension to point cloud processing will lead to artifacts such as shape shrinkage and non-uniform distribution of points. To address these issues, we propose a new point position updating formulation and adopt a multi-normal strategy to overcome sharp edge shrinkage. Compared with other state-of-the-art denoising methods, our approach is more robust in removing small-scale geometric features while retaining large-scale structures. Even compared to its mesh counterpart, our method exhibits superiority in preventing large-scale sharp structures from severe distortion. Finally, a variety of experiments demonstrate that our approach shows its advantages in geometric feature removal against previous methods.