Feature-preserving mesh denoising based on guided normal filtering

Abstract

Mesh denoising is important to improve the quality of the geometry surface acquired by 3D scanning devices. This paper proposes a feature-preserving denoising framework. By classifying the faces into feature and non-feature faces, we use joint bilateral filtering and partial neighborhood filtering to deal with the face normals these two kinds of faces. Experimental results show that our method outperforms the existing methods and achieves higher quality results on the geometry feature.