Texturing and inpainting a complete tubular 3D object reconstructed from partial views

Publication date: Available online 22 May 2018
Source:Computers & Graphics
Author(s): Julien Fayer, Bastien Durix, Simone Gasparini, Géraldine Morin
We present a novel approach to texture 3D tubular objects reconstructed from partial views. Starting from few images of the object, we rely on a 3D reconstruction approach that provides a representation of the object based on a composition of several parametric surfaces, more specifically canal surfaces. Such representation enables a complete reconstruction of the object even for the parts that are hidden or not visible by the input image. The proposed texturing method maps the input images on the parametric surface of the object and complete parts of the surface not visible in any image through an inpainting process. In particular, we first propose a method to select, for each 3D canal surface, the most suitable images and fuse them together for texturing the surface. This process is regulated by a confidence criterion that selects images based on their position and orientation w.r.t. the surface. We also introduce a global method to fuse the images taking into account their exposure difference. Finally, we propose some methods to complete or inpaint the texture in the hidden parts of the surface according to the type of the texture.

Full article: Texturing and inpainting a complete tubular 3D object reconstructed from partial views

Real-time field aligned stripe patterns

Publication date: Available online 22 May 2018
Source:Computers & Graphics
Author(s): Nils Lichtenberg, Noeska Smit, Christian Hansen, Kai Lawonn
In this paper, we present a parameterization technique that can be applied to surface meshes in real-time without time-consuming preprocessing steps. The parameterization is suitable for the display of (un-)oriented patterns and texture patches, and to sample a surface in a periodic fashion. The method is inspired by existing work that solves a global optimization problem to generate a continuous stripe pattern on the surface, from which texture coordinates can be derived. We propose a local optimization approach that is suitable for parallel execution on the GPU, which drastically reduces computation time. With this, we achieve on-the-fly texturing of 3D, medium-sized (up to 70 k vertices) surface meshes. The algorithm takes a tangent vector field as input and aligns the texture coordinates to it. Our technique achieves real-time parameterization of the surface meshes by employing a parallelizable local search algorithm that converges to a local minimum in a few iterations. The calculation in real-time allows for live parameter updates and determination of varying texture coordinates. Furthermore, the method can handle non-manifold meshes. The technique is useful in various applications, e.g., biomedical visualization and flow visualization. We highlight our method’s potential by providing usage scenarios for several applications.

Full article: Real-time field aligned stripe patterns

Edge contraction in persistence-generated discrete Morse vector fields

Publication date: Available online 16 May 2018
Source:Computers & Graphics
Author(s): Tamal K. Dey, Ryan Slechta
Recently, discrete Morse vector fields have been shown to be useful in various applications. Analogous to the simplification of large meshes using edge contractions, one may want to simplify the cell complex K on which a discrete Morse vector field V(K) is defined. To this end, we define a gradient aware edge contraction operator for triangulated 2-manifolds with the following guarantee. If V(K) was generated by a specific persistence-based method, then the vector field that results from our contraction operator is exactly the same as the vector field produced by applying the same persistence-based method to the contracted complex. An implication of this result is that local operations on V(K) are sufficient to produce the persistence-based vector field on the contracted complex. Furthermore, our experiments show that the structure of the vector field is largely preserved by our operator. For example, 1-unstable manifolds remain largely unaffected by the contraction. This suggests that for some applications of discrete Morse theory, it is sufficient to use a contracted complex.

Full article: Edge contraction in persistence-generated discrete Morse vector fields

Automated Outdoor Depth-Map Generation and Alignment

Publication date: Available online 16 May 2018
Source:Computers & Graphics
Author(s): Martin Čadík, Daniel Sýkora, Sungkil Lee
Image enhancement tasks can highly benefit from depth information, but the direct estimation of outdoor depth maps is difficult due to vast object distances. This paper presents a fully automatic framework for model-based generation of outdoor depth maps and its applications to image enhancements. We leverage 3D terrain models and camera pose estimation techniques to render approximate depth maps without resorting to manual alignment. Potential local misalignments, resulting from insufficient model details and rough registrations, are eliminated with our novel free-form warping. We first align synthetic depth edges with photo edges using the as-rigid-as-possible image registration and further refine the shape of the edges using the tight trimap-based alpha matting. The resulting synthetic depth maps are accurate, calibrated in the absolute distance. We demonstrate their benefit in image enhancement techniques including reblurring, depth-of-field simulation, haze removal, and guided texture synthesis.

Full article: Automated Outdoor Depth-Map Generation and Alignment

Surface reconstruction of incomplete datasets: A novel Poisson surface approach based on CSRBF

Publication date: Available online 17 May 2018
Source:Computers & Graphics
Author(s): Jules Morel, Alexandra Bac, Cédric Véga
This paper introduces a novel surface reconstruction method based on unorganized point clouds, which focuses on offering complete and closed mesh models of partially sampled object surfaces. To accomplish this task, our approach builds upon a known a priori model that coarsely describes the scanned object to guide the modeling of the shape based on heavily occluded point clouds. In the region of space visible to the scanner, we retrieve the surface by following the resolution of a Poisson problem: the surface is modeled as the zero level-set of an implicit function whose gradient is the closest to the vector field induced by the 3D sample normals. In the occluded region of space, we consider the a priori model as a sufficiently accurate descriptor of the shape. Both models, which are expressed in the same basis of compactly supported radial functions to ensure computation and memory efficiency, are then blended to obtain a closed model of the scanned object. Our method is finally tested on traditional testing datasets to assess its accuracy and on simulated terrestrial LiDAR scanning (TLS) point clouds of trees to assess its ability to handle complex shapes with occlusions.

Full article: Surface reconstruction of incomplete datasets: A novel Poisson surface approach based on CSRBF

Deep style estimator for 3D indoor object collection organization and scene synthesis

Publication date: Available online 17 May 2018
Source:Computers & Graphics
Author(s): Xiaotian Wang, Bin Zhou, Yu Zhang, Yifan Zhao
Estimating the style compatibility between a pair of cross-category 3D indoor objects has received wide interests from the field of computer graphics in these years. Many previous works solve this task by extracting and analyzing the style-aware structures or elements from the input 3D models. In this paper, we propose a novel approach to solve this task by training a deep neural network to quantitatively assign a compatibility score between arbitrary pair of cross-category 3D objects. By entirely learning from raw data, the trained network is able to capture various compatibility conditions influenced by global style features, such as ergonomics and object category relation. The proposed deep estimator is generally robust and can facilitate various high-level tasks. We first show its application for object collection organization. After that, we show how layout-guided, style-consistent object retrieval for indoor scene synthesis can be achieved by integrating pairwise style estimations into a novel submodular formulation. Our experiments demonstrate the usability of the proposed approach, demonstrating results superior than previous works and even comparable with suggestions made by human observers.

Full article: Deep style estimator for 3D indoor object collection organization and scene synthesis

Sphere-based cut construction for planar parameterizations

Publication date: Available online 18 May 2018
Source:Computers & Graphics
Author(s): Shuangming Chai, Xiao-Ming Fu, Xin Hu, Yang Yang, Ligang Liu
We present a novel algorithm to compute high-quality cuts for generating low isometric distortion planar parameterizations. Based on the observation that conformal spherical and planar parameterizations have similar distortion distributions at the extrusive areas that lead to high isometric distortions, our method utilizes a spherical parameterization of the input mesh to guide the cut construction. After parameterizing the input mesh onto a sphere as conformal as possible, a hierarchical clustering of the divisive type is conducted on the sphere to find high isometric distortion regions, where high isometric distortion may also be introduced in the planar parameterization and which are connected to define a cut. Compared with previous methods, this approach can generate better cuts, resulting in lower isometric distortions. We demonstrate the efficacy and practical robustness of our method on a data set of over 5000 meshes, which are parameterized with low isometric distortion by two existing parameterization approaches.

Full article: Sphere-based cut construction for planar parameterizations

Multi-sided Bézier surfaces over concave polygonal domains

Publication date: Available online 18 May 2018
Source:Computers & Graphics
Author(s): Péter Salvi, Tamás Várady
A new multi-sided, control point based surface representation is introduced, based on the Generalized Bézier patch [1]. While the original surface is based on convex polygonal domains and a specific, uniform arrangement of control points, the new construction permits domains with concave angles and supports a more general control point structure, where independent “half-Bézier” interpolants, or ribbons, are blended together. The ribbons may have arbitrary degrees along the boundary and also in the cross-derivative direction; the related control points ensure tangent- or curvature-continuous connection to adjacent quadrilateral Bézier patches and permit shape editing and optimization, when needed. The surface comprises four components: (i) a concave domain generated from a 3D loop of boundary edges, (ii) half-Bézier interpolants, (iii) parameterizations that cover the full domain for each interpolant, and (iv) blending functions that guarantee both Bézier-like behavior along the boundaries and a smooth, C -continuous composition in the interior of the patch. Editing concave Bézier patches using additional control points is also discussed. A few interesting test examples illustrate the benefits of the method.

Full article: Multi-sided Bézier surfaces over concave polygonal domains

Propagated Mesh Normal Filtering

Publication date: Available online 18 May 2018
Source:Computers & Graphics
Author(s): Bin Liu, Junjie Cao, Weiming Wang, Ning Ma, Bo Li, Ligang Liu, Xiuping Liu
Weighted average is one of the most common strategies used in various mesh filters, and its performance depends on the weight design. When computing the weight between the current face and one of its neighbours, existing methods consider only properties of the two faces, such as positions and normals. Although they generate some convincing results, they definitely tend to suffer from cross-region mixing. For example, assigning such a large weight between two nearby faces separated by some feature edges, even when their properties are close, will damage the local structure. In this paper, we present a novel mesh filter model, named as Propagated Mesh Normal Filtering. It estimates the weight between the current face and its neighbours based on the integral of two kinds of face normal differences along the geodesic path, connecting them. Therefore, prominent features are better preserved when removing noises or textures. Furthermore, in view of the sparseness of large normal difference for most of geometry shapes, the L 1 norm is employed when integrating to further improve the filter. Experiments illustrate the enhanced efficacy of our propagated filter comparing with state-of-the-art methods.

Full article: Propagated Mesh Normal Filtering

2D Skeleton Extraction Based on Heat Equation

Publication date: Available online 18 May 2018
Source:Computers & Graphics
Author(s): Fengyi Gao, Guangshun Wei, Shiqing Xin, Shanshan Gao, Yuanfeng Zhou
Object skeleton is a useful geometric tool for shape analysis tasks. It encodes the topological structure of the primitive shape and preserves a geometric cue as well. Skeletonization is a seemingly simple but difficult research problem. A desirable skeletonization algorithm has to be topologically complete, insensitive to boundary noise, without any branch pruning and free of user-specified parameters. In this paper, we propose a novel method based on the heat equation to extract skeletons assuming that the input is a binary image. Based on the connection between the skeleton and ridge lines, we use the smooth heat diffusion height surface to approximate the real distance field, and then extract the ridge lines of the surface as the output skeleton. Different from the existing approaches, the proposed method in this paper has the above-mentioned excellent properties and can capture the clean and stable skeleton directly. Extensive experimental results show that the new approach can yield better skeletons than the state-of-the-art.

Full article: 2D Skeleton Extraction Based on Heat Equation