Adaptive Moving Least Squares for Smoothing Point Clouds

Jianping Cai, Zhongke Wu, Xingce Wang, Mingquan Zhou, and Pengfei Xu


Point set surface, Moving least squares (MLS), Point clouds smoothing


The moving least squares (MLS) surface, defined algorithmically as the output of a particular meshless reconstruction, has been widely used for modeling and rendering with point clouds. In this paper, a modified method based on MLS surface is proposed to smooth some noisy point cloud models. Nearest-neighbours search results of k-d tree are further processed where there are holes or close-by sheet surfaces; Erroneous unsigned normal directions estimation by PCA analysis at sharp corners are corrected; Local reference domain of MLS projection is processed by removing some undesired region at sharp corners. With these improvements, the robustness of MLS algorithm is improved, and noise are better suppressed while sharp features are retained.

Important Links:

Go Back