Adaptive and propagated mesh filtering

Abstract

In this paper, we present an adaptive and propagated mesh filtering algorithm, which utilizes the common two-stage mesh denoising architecture. In the first stage, face normals are filtered by calculating cumulative differences of two kinds of properties along the geodesic paths connecting centroids of face neighbors. Compared with the well-known bilateral strategy and its variants, our strategy respects the spatial structure of mesh itself through using intrinsic metrics. Furthermore, the parameters in the weight strategy of our algorithm can be fully determined adaptively, which avoids the dilemma that previous methods use trial and error to get satisfactory denoised results. In the second stage, mean curvature flow theorem is used to update the vertex positions, namely, the direction of vertex update is adjusted to a linear combination between the direction of differential coordinates and the descending direction given by least squares, which improves the quality of meshes. Extensive experiments demonstrate the effectiveness of our method in preserving latent geometric features compared with the state-of-the-art approaches. • Face normals are filtered by calculating cumulative differences of two kinds of properties along the geodesic paths connecting centroids of face neighbors and the parameters in the weight strategy of our algorithm can be fully determined adaptively. • Compared with the well-known bilateral strategy and its variants, our strategy respects the spatial structure of mesh itself through using intrinsic metrics. • In addition, we use mean curvature flow theorem to update vertex positions, which improves the quality of mesh.