Submitted to SIAM Conference on Geometric Design and Computing. Also Extended Abstract.
Abstract. We present a new method for point cloud denoising. We introduce a robust smoothing operator Q(p)=p+t'n', inspired in moving least squares and M-estimators robust statistics theory. Our algorithm can be seen as a generalization and improvement of the Fleishman et al algorithm for mesh denoising. We also extend the strategies for mesh denoising to point clouds with some improvements to handle oversmoothing and input models with thin regions.
Keywords: point cloud; moving least squares; M-estimators; k-nearest neighbors.
|
|
|
|
|
| Noisy Input Model | Fleishman et al Method | Jones et al Method | Our Method |
|
|
|
|
| Noisy Input Model | Fleishman et al Method | Jones et al Method | Our Method |
|
|
|
|
| Noisy Input Model | Fleishman et al Method | Jones et al Method | Our Method |