Ulrich Schwanecke

Feature sensitive surface extraction from volume data

The representation of geometric objects based on volumetric data structures has advantages in many geometry processing applications that require, e.g., fast surface interrogation or boolean operations such as intersection and union. However, surface based algorithms like shape optimization (fairing) or freeform modeling often need a topological manifold representation where neighborhood information within the surface is explicitly available. Consequently, it is necessary to find effective conversion algorithms to generate explicit surface descriptions for the geometry which is implicitly defined by a volumetric data set. Since volume data is usually sampled on a regular grid with a given step width, we often observe severe alias artifacts at sharp features on the extracted surfaces. In this paper we present a new technique for surface extraction that performs feature sensitive sampling and thus reduces these alias effects while keeping the simple algorithmic structure of the standard Marching Cubes algorithm. We demonstrate the effectiveness of the new technique with a number of application examples ranging from CSG modeling and simulation to surface reconstruction and remeshing of polygonal models.

Artefacts in CBCT: a review

Artefacts are common in todays cone beam CT (CBCT). They are induced by discrepancies between the mathematical modelling and the actual physical imaging process. Since artefacts may interfere with the diagnostic process performed on CBCT data sets, every user should be aware of their presence. This article aims to discuss the most prominent artefacts identified in the scientific literature and review the existing knowledge on these artefacts. We also briefly review the basic three-dimensional (3D) reconstruction concept applied by todays CBCT scanners, as all artefacts are more or less directly related to it.

Interactive visualization for opportunistic exploration of large document collections

Finding relevant information in a large and comprehensive collection of cross-referenced documents like Wikipedia usually requires a quite accurate idea where to look for the pieces of data being sought. A user might not yet have enough domain-specific knowledge to form a precise search query to get the desired result on the first try. Another problem arises from the usually highly cross-referenced structure of such document collections. When researching a subject, users usually follow some references to get additional information not covered by a single document. With each document, more opportunities to navigate are added and the structure and relations of the visited documents gets harder to understand. This paper describes the interactive visualization Wivi which enables users to intuitively navigate Wikipedia by visualizing the structure of visited articles and emphasizing relevant other topics. Combining this visualization with a view of the current article results in a custom browser specially adapted for exploring large information networks. By visualizing the potential paths that could be taken, users are invited to read up on subjects relevant to the current point of focus and thus opportunistically finding relevant information. Results from a user study indicate that this visual navigation can be easily used and understood. A majority of the participants of the study stated that this method of exploration supports them finding information in Wikipedia.

Analysis and design of Hermite subdivision schemes

Starting from an initial sequence of Hermite elements, a Hermite subdivision scheme recursively generates finer sequences of Hermite elements which are associated with the dyadic points. With the help of the interpolating splines that can be associated with the Hermite elements, we analyze the smoothness of the limit curves generated by Hermite subdivision schemes of arbitrary order, including non-interpolatory ones. After presenting these theoretical results, we describe two new families of Hermite subdivision schemes.

Automated detection of patient movement during a CBCT scan based on the projection data

OBJECTIVES To develop an automated procedure to detect patient motion on the projection images acquired during a cone beam computed tomography (CBCT) scan and to evaluate the methods feasibility on small real-world CBCT images in relation to visual assessment. METHODS Based on optical flow theory, software was developed using the sequence of the projection images of a CBCT machine for automated detection of patient motion. Averaged acceleration vectors were used as measurement data and compared with visual assessment of the projection images displayed as video. Seventy-nine CBCT data sets (small field-of-view: 40 mm) from our patient database were selected in a sequential fashion and evaluated with the software. RESULTS 10 out of 79 (13%) were allocated to a patient movement. A threshold of 0.4 pixel/frame transition was empirically determined as indicating motion by visual assessment of the image sequence. Relative to this standard of reference, the software reached 80% sensitivity versus 67% specificity. CONCLUSIONS Optical flow seems to be an efficient concept for automated detection of patient motion on the projection images acquired during a CBCT scan.

GPU-Based Volume Reconstruction from Very Few Arbitrarily Aligned X-Ray Images

This paper presents a three-dimensional GPU-accelerated algebraic reconstruction method in a few-projection cone-beam setting with arbitrary acquisition geometry. To achieve artifact-reduced reconstructions in the challenging case of unconstrained geometry and extremely limited input data, we use linear methods and an artifact-avoiding projection algorithm to provide high reconstruction quality. We apply the conjugate gradient method in the linear case of Tikhonov regularization and the two-point-step-size gradient method in the nonlinear case of total variation regularization to solve the system of equations. By taking advantage of modern graphics hardware we achieve acceleration of up to two orders of magnitude over classical CPU implementations.

Real-Time Monocular Segmentation and Pose Tracking of Multiple Objects

We present a real-time system capable of segmenting multiple 3D objects and tracking their pose using a single RGB camera, based on prior shape knowledge. The proposed method uses twist-coordinates for pose parametrization and a pixel-wise second-order optimization approach which lead to major improvements in terms of tracking robustness, especially in cases of fast motion and scale changes, compared to previous region-based approaches. Our implementation runs at about 50–100 Hz on a commodity laptop when tracking a single object without relying on GPGPU computations. We compare our method to the current state of the art in various experiments involving challenging motion sequences and different complex objects.

Blind Image Deconvolution of Linear Motion Blur

We present an efficient method to deblur images for information recognition. The method is successfully applied directly on mobile devices as a preprocessing phase to images of barcodes. Our main contribution is the fast identifaction of blur length and blur angle in the frequency domain by an adapted radon transform. As a result, the barcode recognition rate of the deblurred images has been increased significantly.

Metal artifact reduction in x-ray computed tomography: Inpainting versus missing value

A comparison of algorithms for reduction of metal artifacts in x-ray cone beam computed tomography (CBCT) is presented. In the context of algebraic reconstruction techniques (ART) several inpainting algorithms in the image domain are evaluated against missing data strategies. A GPU-based iterative framework is employed for a meaningful comparison of both. Simulation results from an extended Shepp-Logan phantom and real world dental data are given.

High-Speed and Robust Monocular Tracking

In this paper, we present a system for high-speed robust monocular tracking (HSRM-Tracking) of active markers. The proposed algorithm robustly and accurately tracks multiple markers at full framerate of current high-speed cameras. For this, we have developed a novel, nearly co-planar marker pattern that can be identified without initialization or incremental tracking. The pattern also encodes a unique ID to identify different markers. The individual markers are calibrated semi-automatically, thus no time-consuming and error-prone manual measurement is needed. Finally we show that the minimal spatial structure of the marker can be used to robustly avoid pose ambiguities even at large distances to the camera. This allows us to measure the pose of each individual marker with high accuracy in a vast area.