Jochen Süßmuth

Automatic reconstruction of personalized avatars from 3D face scans

We present a simple algorithm for computing a high‐quality personalized avatar from a single color image and the corresponding depth map which have been captured by Microsofts Kinect sensor. Due to the low market price of our hardware setup, 3D face scanning becomes feasible for home use. The proposed algorithm combines the advantages of robust non‐rigid registration and fitting of a morphable face model. We obtain a high‐quality reconstruction of the facial geometry and texture along with one‐to‐one correspondences with our generic face model. This representation allows for a wide range of further applications such as facial animation or manipulation. Our algorithm has proven to be very robust. Since it does not require any user interaction, even non‐expert users can easily create their own personalized avatars. Copyright

Reconstructing animated meshes from time-varying point clouds

In this paper, we describe a novel approach for the reconstruction of animated meshes from a series of time‐deforming point clouds. Given a set of unordered point clouds that have been captured by a fast 3‐D scanner, our algorithm is able to compute coherent meshes which approximate the input data at arbitrary time instances. Our method is based on the computation of an implicit function in ℝ4 that approximates the time‐space surface of the time‐varying point cloud. We then use the four‐dimensional implicit function to reconstruct a polygonal model for the first time‐step. By sliding this template mesh along the time‐space surface in an as‐rigid‐as‐possible manner, we obtain reconstructions for further time‐steps which have the same connectivity as the previously extracted mesh while recovering rigid motion exactly.

On floating-point normal vectors

In this paper we analyze normal vector representations. We derive the error of the most widely used representation, namely 3D floating‐point normal vectors. Based on this analysis, we show that, in theory, the discretization error inherent to single precision floating‐point normals can be achieved by 250.2 uniformly distributed normals, addressable by 51 bits. We review common sphere parameterizations and show that octahedron normal vectors perform best: they are fast and stable to compute, have a controllable error, and require only 1 bit more than the theoretical optimal discretization with the same error.

Surface Reconstruction Based on Hierarchical Floating Radial Basis Functions

In this paper we address the problem of optimal centre placement for scattered data approximation using radial basis functions (RBFs) by introducing the concept of floating centres. Given an initial least‐squares solution, we optimize the positions and the weights of the RBF centres by minimizing a non‐linear error function. By optimizing the centre positions, we obtain better approximations with a lower number of centres, which improves the numerical stability of the fitting procedure. We combine the non‐linear RBF fitting with a hierarchical domain decomposition technique. This provides a powerful tool for surface reconstruction from oriented point samples. By directly incorporating point normal vectors into the optimization process, we avoid the use of off‐surface points which results in less computational overhead and reduces undesired surface artefacts. We demonstrate that the proposed surface reconstruction technique is as robust as recent methods, which compute the indicator function of the solid described by the point samples. In contrast to indicator function based methods, our method computes a global distance field that can directly be used for shape registration.

Ridge based curve and surface reconstruction

This paper presents a new method for reconstructing curves and surfaces from unstructured point clouds, allowing for noise in the data as well as inhomogeneous distribution of the point set. It is based on the observation that the curve/surface is located where locally the point cloud has highest density. This idea is pursued by a differential geometric analysis of a smoothed version of the density function. More precisely we detect ridges of this function and have to single out the relevant parts. An efficient implementation of this approach evaluates the differential geometric quantities on a regular grid, performs local analysis and finally recovers the curve/surface by an isoline extraction or a marching cubes algorithm respectively. Compared to existing surface reconstruction procedures, this approach works well for noisy data and for data with strongly varying sampling rate. Thus it can be applied successfully to reconstruct surface geometry from time-of-flight data, overlapping registered point clouds and point clouds obtained by feature tracking from video streams. Corresponding examples are presented to demonstrate the advantages of our method.

GPU based ARAP Deformation using Volumetric Lattices

We present a novel lattice based direct manipulation paradigm (LARAP) for mesh editing that decouples the runtime complexity from the mesh’s geometric complexity. Since our non-linear optimization is based on the ARAP paradigm, it is very fast and can be easily implemented. Our proxy geometry automatically introduces volumeawareness into the optimization problem, leading to more natural deformations. Since we compute how the space surrounding an object has to be deformed to satisfy a set of user-constraints, we can even handle models with disconnected parts. We analyze the bottlenecks of the presented approach and propose a data-parallel multiresolution implementation on the GPU, which allows to pose even high-quality meshes consisting of millions of triangles in real-time.

Multiresolution attributes for tessellated meshes

We present a novel representation for storing sub-triangle signals, such as colors, normals, or displacements directly with the triangle mesh. Signal samples are stored as guided by hardware-tessellation patterns. Thus, we can directly render from our representation by assigning signal samples to attributes of vertices generated by the hardware tessellator. Contrary to texture mapping, our approach does not require any atlas generation, chartification, or uv-unwrapping. Thus, it does not suffer from texture-related artifacts, such as discontinuities across chart boundaries or distortion. Moreover, our approach allows specifying the optimal sampling rate adaptively on a per triangle basis, resulting in significant memory savings for most signal types. We propose a signal optimal approach for converting arbitrary signals, including existing assets with textures or mesh colors, into our representation. Further, we provide efficient algorithms for mip-mapping, bi- and tri-linear interpolation directly in our representation. Our approach is optimally suited for displacement mapping: it automatically generates crack-free, view-dependent displacement mapped models enabling continuous level-of-detail.

Animation transplantation

Realistic character animation requires elaborate rigging built on top of high quality 3D models. Sophisticated anatomically based rigs are often the choice of visual effect studios where life-like animation of CG characters is the primary objective. However, rigging a character with a muscular-skeletal system is very involving and time-consuming process, even for professionals. Although, there have been recent research efforts to automate either all or some parts of the rigging process, the complexity of anatomically based rigging nonetheless opens up new research challenges. We propose a new method to automate anatomically based rigging that transfers an existing rig of one character to another. The method is based on a data interpolation in the surface and volume domain, where various rigging elements can be transferred between different models. As it only requires a small number of corresponding input feature points, users can produce highly detailed rigs for a variety of desired character with ease. Copyright

Color-encoded distance visualization of cranial nerve-vessel contacts

PurposeVisualization of pathological contact between cranial nerves and vascular structures at the surface of the brainstem is important for diagnosis and treatment of neurovascular compression (NVC) syndromes. We developed a method for improved visualization of this abnormality.MethodsDistance fields were computed using preoperative MRI scans of individuals with NVC syndromes to support the topological representation of brainstem surface structures with quantitative information. Polygonal models of arteries, cranial nerves and the brainstem were generated using segmented T2 weighted MR data. After color-coding the polygonal models with the respective distances, enhanced color visualization of vessel-nerve locations with possible contacts was achieved.ResultsThe proposed method was implemented and applied to surgical planning in a dozen cases of NVC syndrome. Two selected cases were chosen to demonstrate the feasibility and subjective improvement provided by our visualization technique. Expert neurosurgeons found the improvement valuable and useful for these cases.ConclusionColor-encoded distance information significantly improves the perceptibility of potential nerve-vessel contacts. This method contributes to a better understanding of the complex anatomical situation at the surface of the brainstem and assists in planning of surgery.

Nutzenpotenziale maßgetreuer 3D-Avatare aus Low-Cost-Bodyscannern

Der Einsatz realitätsnaher dreidimensionaler Modelle des menschlichen Körpers (sog. Avatare) wird seit Mitte der 90er-Jahre für unterschiedlichste Anwendungen diskutiert. Eine kommerzielle Umsetzung scheiterte in der Vergangenheit jedoch oft an den Kosten von 3D-Scannern. Neue technologische Innovationen, insbesondere die von der Unterhaltungsindustrie forcierte Entwicklung preiswerter Tiefensensoren, führen nun dazu, dass 3D-Scanner heute zu einem Bruchteil der Kosten entwickelt werden können. So entsteht die Möglichkeit für den breiten Einsatz von Avataren in aufgabenorientierten und nicht aufgabenorientierten Kontexten und damit ein enormes betriebswirtschaftliches Potenzial.