Jörg Peters
The note shows how $G^k$ (geometrically continuous surface) constructions yield $C^k$ iso-geometric elements also at irregular quad mesh points where three or more than four elements come together.
Kestutis Karciauskas, Daniele Panozzo, Jörg Peters
This paper develops a new way to create smooth piecewise polynomial free-form spline surfaces from quad- meshes that include T-junctions, where surface strips start or terminate. All mesh nodes can be interpreted as control points of geometrically-smooth, piecewise polynomials that we call GT-splines. GT-splines are B-spline-like and cover T-junctions by two or four patches of degree bi-4. They complement multi-sided surface constructions in generating free-form surfaces with adaptive layout. Since GT-splines do not require a global coordination of knot intervals, GT-constructions are easy to deploy and can provide smooth surfaces with T-junctions where T-splines can not have a smooth parameterization. GT-constructions display a uniform highlight line distribution on input meshes where alternatives, such as Catmull-Clark subdivision, exhibit oscillations.
Jörg Peters
Splines can be constructed by convolving the indicator function of a cell whose shifts tessellate $\R^k$. This paper presents simple, non-algebraic criteria that imply that, for regular shift-invariant tessellations, only a small subset of such spline families yield nested spaces: primarily the well-known tensor-product and box splines. Among the many non-refinable constructions are hex-splines and their generalization to the Voronoi cells of non-Cartesian root lattices.
Dang-Manh Nguyen, Jörg Peters
Convolving the output of Discontinuous Galerkin (DG) computations using spline filters can improve both smoothness and accuracy of the output. At domain boundaries, these filters have to be one-sided. Recently, position-dependent smoothness-increasing accuracy-preserving (PSIAC) filters were shown to be a superset of the top-of-the-line one-sided RLKV and SRV filters. Since PSIAC filters can be formulated symbolically, convolution with PSIAC filters reduces to a stable inner product with the DG output and hence provides a stable and efficient implementation. The paper focuses on the remarkable fact that, for the canonical linear or non-linear hyperbolic test equation, new piecewise constant PSIAC filters of small support outperform the top-of-the-line boundary filters in the literature. These least-degree filters reduce the error at the boundaries to less than the error even of the symmetric filters of the same support that are applied in the interior . Due to their simplicity, and since this least degree filter has an exact symbolic form, convolution is stable as well as efficient and derivatives of the convolved output are easy to compute.
Jörg Peters
The Discontinuous Galerkin (DG) method applied to hyperbolic differential equations outputs weakly-linked polynomial pieces. Post-processing these pieces by Smoothness-Increasing Accuracy-Conserving (SIAC) convolution with B-splines can improve the accuracy of the output and yield superconvergence. SIAC convolution is considered optimal if the SIAC kernels, in the form of a linear combinations of B-splines of degree d, reproduce polynomials of degree 2d. This paper derives simple formulas for computing the optimal SIAC spline coefficients.
Kyle Shih-Huang Lo, Jörg Peters, Eric Spellman
Accurate completion and denoising of roof height maps are crucial to reconstructing high-quality 3D buildings. Repairing sparse points can enhance low-cost sensor use and reduce UAV flight overlap. RoofDiffusion is a new end-to-end self-supervised diffusion technique for robustly completing, in particular difficult, roof height maps. RoofDiffusion leverages widely-available curated footprints and can so handle up to 99\% point sparsity and 80\% roof area occlusion (regional incompleteness). A variant, No-FP RoofDiffusion, simultaneously predicts building footprints and heights. Both quantitatively outperform state-of-the-art unguided depth completion and representative inpainting methods for Digital Elevation Models (DEM), on both a roof-specific benchmark and the BuildingNet dataset. Qualitative assessments show the effectiveness of RoofDiffusion for datasets with real-world scans including AHN3, Dales3D, and USGS 3DEP LiDAR. Tested with the leading City3D algorithm, preprocessing height maps with RoofDiffusion noticeably improves 3D building reconstruction. RoofDiffusion is complemented by a new dataset of 13k complex roof geometries, focusing on long-tail issues in remote sensing; a novel simulation of tree occlusion; and a wide variety of large-area roof cut-outs for data augmentation and benchmarking.
Jörg Peters
Lower bounds on the generation of smooth bi-cubic surfaces imply that geometrically smooth ($G^1$) constructions need to satisfy conditions on the connectivity and layout. In particular, quadrilateral meshes of arbitrary topology can not in general be covered with $G^1$ -connected Bézier patches of bi-degree 3 using the layout proposed in [ASC17]. This paper analyzes whether the pre-refinement of the input mesh by repeated Doo-Sabin subdivision proposed in that paper yields an exception.
Kestutis Karciauskas, Jörg Peters
Converting quad meshes to smooth manifolds, guided subdivision offers a way to combine the good highlight line distributions of recent G-spline constructions with the refinability of subdivision surfaces. Specifically, we present a C2 subdivision algorithm of polynomial degree bi-6 and a curvature bounded algorithm of degree bi-5. We prove that the common eigen-structure of this class of subdivision algorithms is determined by their guide and demonstrate that the eigenspectrum (speed of contraction) can be adjusted without harming the shape.
Dang-Manh Nguyen, Jörg Peters
Convolving the output of Discontinuous Galerkin (DG) computations with symmetric Smoothness-Increasing Accuracy-Conserving (SIAC) filters can improve both smoothness and accuracy. To extend convolution to the boundaries, several one-sided spline filters have recently been developed. We interpret these filters as instances of a general class of position-dependent spline filters that we abbreviate as PSIAC filters. These filters may have a non-uniform knot sequence and may leave out some B-splines of the sequence. For general position-dependent filters, we prove that rational knot sequences result in rational filter coefficients. We derive symbolic expressions for prototype knot sequences, typically integer sequences that may include repeated entries and corresponding B-splines, some of which may be skipped. Filters for shifted or scaled knot sequences are easily derived from these prototype filters so that a single filter can be re-used in different locations and at different scales. Moreover, the convolution itself reduces to executing a single dot product making it more stable and efficient than the existing approaches based on numerical integration. The construction is demonstrated for several established and one new boundary filter.