Guided point cloud denoising via sharp feature skeletons.

Feature-preserving filtering of noisy point clouds plays a fundamental role in geometric processing. Though the guided filter is known to be a powerful tool for edge-aware image processing and mesh denoising, extending it to point clouds is not a trivial task due to the difficulty of defining a piecewise smooth normal field on point clouds with sharp features. Our key idea to address the issue is to assign feature points with multiple normals according to their feature type. Specifically, our approach consists of four stages. It first screens out candidate feature points according to normal variation and then employs the $$l_1$$ -medial skeleton to extract a sharp feature structure. Following that, multiple normals are computed for each feature point by using k-means clustering. It then computes the guidance normals by using a k-nearest neighbor patch whose normals are most consistent. Point positions are finally updated according to the filtered normals. A variety of experiments suggest that our approach can robustly filter out high level of noise while keeping the important geometric features intact. [ABSTRACT FROM AUTHOR]