Recognising Algebraic Surfaces from Two Outlines

Photographic outlines of 3 dimensional solids are robust and rich in information useful for surface reconstruction. This paper studies algebraic surfaces viewed from 2 cameras with known intrinsic and extrinsic parameters. It has been known for some time that for a degree d=2 (quadric) algebraic surface there is a 1-parameter family of surfaces that reproduce the outlines. When the algebraic surface has degree d>2, we prove a new result: that with known camera geometry it is possible to completely reconstruct an algebraic surface from 2 outlines i.e. the coefficients of its defining polynomial can be determined in a known coordinate frame. The proof exploits the existence of frontier points, which are calculable from the outlines. Examples and experiments are presented to demonstrate the theory and possible applications.

Neural Representation of Open Surfaces

Neural implicit surfaces have emerged as an effective, learnable representation for shapes of arbitrary topology. However, representing open surfaces remains a challenge. Different methods, such as unsigned distance fields (UDF), have been proposed to tackle this issue, but a general solution remains elusive. The generalized winding number (GWN), which is often used to distinguish interior points from exterior points of 3D shapes, is arguably the most promising approach. The GWN changes smoothly in regions where there is a hole in the surface, but it is discontinuous at points on the surface. Effectively, this means that it can be used in lieu of an implicit surface representation while providing information about holes, but, unfortunately, it does not provide information about the distance to the surface necessary for e.g. ray tracing, and special care must be taken when implementing surface reconstruction. Therefore, we introduce the semi-signed distance field (SSDF) representation which comprises both the GWN and the surface distance. We compare the GWN and SSDF representations for the applications of surface reconstruction, interpolation, reconstruction from partial data, and latent vector analysis using two very different data sets. We find that both the GWN and SSDF are well suited for neural representation of open surfaces.     

Reconstructing Surfaces and Functions on Surfaces from Unorganized Three-Dimensional Data

Creating a computer model from an existing part is a common problem in reverse engineering. The part might be scanned with a device like the laser range scanner, or points might be measured on its surface with a mechanical probe. Sometimes, not only the spatial location of points, but also some associated physical property can be measured. The problem of automatically reconstructing from this data a topologically consistent and geometrically accurate model of the object and of the sampled scalar field is the subject of this paper. The proposed algorithm can deal with connected, orientable manifolds of unrestricted topological type, given a sufficiently dense and uniform sampling of the object's surface. It is capable of automatically reconstructing both the model and a scalar field over its surface. It uses Delaunay triangulations, Voronoi diagrams, and α-shapes for efficiency of computation and theoretical soundness. It generates a representation of the surface and the field based on Bernstein—Bézier polynomials, with the surface modeled by implicit patches (A-patches), that are guaranteed to be smooth and single-sheeted. Received June 1, 1994; revised February 2, 1995, and August 14, 1995.     

Adaptive Sampling of Implicit Surfaces for Interactive Modelling and Animation

This paper presents a new adaptive sampling method for implicit surfaces that can be used in both interactive modelling and animation. The algorithm samples implicit objects composed of blending primitives and efficiently maintains this sampling over time, even when their topology changes (during fractures and fusions). It provides two complementary modes of immediate visualization: displaying “scales” lying on the surface, or a “primitive-wise” polygonization. The sampling method efficiently avoids unwanted blending between different parts of an object. Moreover, it can be used for partitioning an implicit surface into local bounding boxes that will accelerate collision detection during animation and ray-intersections during final rendering.     

基于隐式曲面的3维树木建模

为了利用隐式曲面构造出光滑拼接的3维树木枝条模型,同时义能避免3维树木建模中枝条曲面的单一性,以增强树木模型局部细节的真实感描述.提出了一种基于隐式曲面的3维树木建模方法,该方法首先将基于骨架的隐式曲面与具有多项式密度分布的直线骨架卷积曲面结合起来进行3维树木建模;然后采用BlobTree结构来组合隐式曲面原型,并用优化融合来消除树木枝条的融合突起;最后用PCM等隐式曲面建模技术来模拟3维树木表面的局部细节,并采用实际树木图片中的颜色概率分布进行纹理填充.实验结果表明,该方法不仅可以重构出具有多样性的光滑3维树木枝条,并能够逼真地生成树权脊梁、树木突起等局部细节特征.     

一种改进的八元树三维目标表示方法

利用八元树表示三维物体是一种十分有效的方法。传统指针表示的八元树结构具有占据内存容量大、节点间关系少等缺点。在分析了物体的空间布局和八元树结构的基础上,文章提出了一种改进的八元树三维表示方法。从医学图像三维重建的实验结果看,提出的方法可行并且在内存存储结构和访问方式等方面优于传统的表示方法。     

Normal Computation for Discrete Surfaces in 3D Space

Associating normal vectors to surfaces is essential for many rendering algorithms. We introduce a new method to compute normals on discrete surfaces in object space. Assuming that the surface separates space locally into two disjoint subsets, each of these subsets contains implicitly information about the surface inclination. Considering one of these subsets in a small neighbourhood of a surface point enables us to derive the surface normal from this set. We show that this leads to exact results for C¹ continuous surfaces in R³. Furthermore, we show that good approximations can be obtained numerically by sampling the considered area. Finally, we derive a method for normal computation on surfaces in discrete space.     

Multiresolution for Algebraic Curves and Surfaces using Wavelets

This paper describes a multiresolution method for implicit curves and surfaces. The method is based on wavelets, and is able to simplify the topology. The implicit curves and surfaces are defined as the zero-valued piece-wise algebraic isosurface of a tensor-product uniform cubic B-spline. A wavelet multiresolution method that deals with uniform cubic B-splines on bounded domains is proposed. In order to handle arbitrary domains the proposed algorithm dynamically adds appropriate control points and deletes them in the synthesis phase.