VMV13

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/1017

Browse

Now showing 1 - 20 of 34

Sampled and Analytic Rasterization
(The Eurographics Association, 2013) Auzinger, Thomas; Wimmer, Michael; Michael Bronstein and Jean Favre and Kai Hormann
In this poster we present an overview of exact anti-aliasing (AA) methods in rasterization. In contrast to the common supersampling approaches for visibility AA (e.g. MSAA) or both visibility and shading AA (e.g. SSAA, decoupled sampling), prefiltering provides the mathematically exact solution to the aliasing problem. Instead of averaging a set a supersamples, the input data is convolved with a suitable low-pass filter before sampling is applied. Recent work showed that for both visibility signals and simple shading models, a closed-form solution to the convolution integrals can be found. As our main contribution, we present a classification of both sample-based and analytic AA approaches for rasterization and analyse their strengths and weaknesses.
Transfusive Weights for Content-Aware Image Manipulation
(The Eurographics Association, 2013) Yücer, Kaan; Sorkine-Hornung, Alexander; Sorkine-Hornung, Olga; Michael Bronstein and Jean Favre and Kai Hormann
Many image editing operations, such as colorization, matting or deformation, can be performed by propagating user-defined sparse constraints (e.g. scribbles) to the rest of the image using content-aware weight functions. Image manipulation has been recently extended to simultaneous editing of multiple images of the same subject or scene by precomputing dense correspondences, where the content-aware weights play a core role in defining the sub-pixel accurate image warps from source to target images. In this paper, we expand the range of applications for content-aware weights to the multi-image setting and improve the quality of the recently proposed weights and the matching framework. We show that multiple images of a subject can be used to refine the content-aware weights, and we propose a customization of the weights to enable easily-controllable interactive depth segmentation and assignment, image matting and deformation transfer, both in single- and multi-image settings.
Closed-Form Hierarchical Finite Element Models for Part-Based Object Detection
(The Eurographics Association, 2013) Rak, Marko; Engel, Karin; Tönnies, Klaus D.; Michael Bronstein and Jean Favre and Kai Hormann
In this work we address part-based object detection under variability of part shapes and spatial relations. Our approach bases on the hierarchical finite element modeling concept of Engel and Tönnies [ET09a, ET09b]. They model object parts by elastic materials, which adapt to image structures via image-derived forces. Spatial part relations are realized through additional layers of elastic material forming an elastic hierarchy. We present a closed-form solution to this concept, reformulating the hierarchical optimization problem into the optimization of a non-hierarchical finite element model. This allows us to apply standard finite element techniques to hierarchical problems and to provide an efficient framework for part-based object detection. We demonstrate our approach at the example of lumbar column detection in magnetic resonance imaging on a data set of 49 subjects. Given a rough model initialization, our approach solved the detection problem reliably in 45 out of 49 cases, showing computation times of only a few seconds per subject.
A New Framework for Fitting Shape Models to Range Scans: Local Statistical Shape Priors Without Correspondences
(The Eurographics Association, 2013) Last, Carsten; Winkelbach, Simon; Wahl, Friedrich M.; Michael Bronstein and Jean Favre and Kai Hormann
Statistical shape models provide an important means in many applications in computer vision and computer graphics. However, the major problems are that the majority of these shape models require dense pointcorrespondences along all training shapes and that a large number of training shapes is needed in order to capture the full amount of intra-class shape variation. In this contribution, we focus on a statistical shape model that can be constructed from a set of training shapes without defining any point-correspondences. Additionally, we show how a local statistical shape model can make better use of the available shape information, greatly reducing the number of required training shapes. Finally, we present a new framework to fit this local statistical shape model without correspondences to range scans that represent incomplete parts of the trained shape class. The fitted model is then used to reproduce a natural-looking approximation of the complete shape.
Interactive Direct Volume Rendering with Many-light Methods and Transmittance Caching
(The Eurographics Association, 2013) Weber, Christoph; Kaplanyan, Anton S.; Stamminger, Marc; Dachsbacher, Carsten; Michael Bronstein and Jean Favre and Kai Hormann
In this paper we present an interactive global illumination method for Direct Volume Rendering (DVR) based on the many-light approach, a class of global illumination methods which gained much interest recently. We extend these methods to handle transfer function and volume density updates efficiently in order to foster ability of interactive volume exploration. Global illumination techniques accounting for all light transport phenomena are typically computationally too expensive for interactive DVR. Many-light methods represent the light transport in a volume by determining a set of virtual light sources whose direct illumination and single scattering to a view ray approximate full global illumination. Our technique reduces computation caused by transfer function changes by recomputing the contribution of these virtual lights, and rescaling or progressively updating their volumetric shadow maps and locations. We discuss these optimizations in the context of DVR and demonstrate their application to interactive rendering.
Audio Resynthesis on the Dancefloor: A Music Structural Approach
(The Eurographics Association, 2013) Tauscher, Jan-Philipp; Wenger, Stephan; Magnor, Marcus; Michael Bronstein and Jean Favre and Kai Hormann
We propose a method for synthesizing a novel soundtrack from an existing musical piece while preserving its structure and continuity from a music theoretical point of view. Existing approaches analyze a musical piece for possible cut points that allow the resynthesis of a novel soundtrack by lining up the source segments according to specified rules but fail to maintain musically correct song progression. Introducing the alignment of rhythmic and harmonic structures during transition point detection, we employ beat tracking as the analysis core component and take the human sound perception into account. Automatic segment rearrangement is improved by employing a novel belief propagation approach that enables user-defined constraints for the output soundtrack, allowing video editors or dance choreographers to tailor a soundtrack to their specific demands.
Non-Sampled Anti-Aliasing
(The Eurographics Association, 2013) Auzinger, Thomas; Musialski, Przemyslaw; Preiner, Reinhold; Wimmer, Michael; Michael Bronstein and Jean Favre and Kai Hormann
In this paper we present a parallel method for high-quality edge anti-aliasing in rasterization. In contrast to traditional graphics hardware methods, which rely on massive oversampling to combat aliasing issues, we evaluate a closed-form solution of the associated prefilter convolution. This enables the use of a wide range of filter functions with arbitrary kernel sizes, as well as general shading methods such as texture mapping or complex illumination models. Due to the use of analytic solutions, our results are exact in the mathematical sense and provide objective ground-truth for other anti-aliasing methods and enable the rigorous comparison of different models and filters. An efficient implementation on general purpose graphics hardware is discussed and several comparisons to existing techniques and of various filter functions are given.
Parallelized Global Brain Tractography
(The Eurographics Association, 2013) Philips, Stefan; Hlawitschka, Mario; Scheuermann, Gerik; Michael Bronstein and Jean Favre and Kai Hormann
Most brain tractography algorithms suffer from lower accuracy, because they use only information in a certain neighborhood and reconstruct the tracts independently. Global brain tractography algorithms compensate the lack of accuracy of those local algorithms in certain areas by optimizing the whole tractogram. The global tractography approach by Reisert et al. showed the best results in the Fiber Cup contest, but the runtime is still a matter for a medical application. In this paper we present the non-trivial parallelization of this global tractography algorithm. The parallelization exploits properties of the algorithm and modifies the algorithm where necessary. We compare the runtimes of the serial and the parallel variant and show that the outcomes of the parallel variant are of the same quality as those of the serial algorithm. The experiments proof also that the parallelization scales well for real world datasets.
A Thin Shell Approach to the Registration of Implicit Surfaces
(The Eurographics Association, 2013) Iglesias, Jose A.; Berkels, Benjamin; Rumpf, Martin; Scherzer, Otmar; Michael Bronstein and Jean Favre and Kai Hormann
Frequently, one aims at the co-registration of geometries described implicitly by images as level sets. This paper proposes a novel shape sensitive approach for the matching of such implicit surfaces. Motivated by physical models of elastic shells a variational approach is proposed, which distinguishes two different types of energy contributions: a membrane energy measuring the rate of tangential distortion when deforming the reference surface into the template surface, and a bending energy reflecting the required amount of bending. The variational model is formulated via a narrow band approach. The built in tangential distortion energy leads to a suitable equidistribution of deformed length and area elements, under the optimal matching deformation, whereas the minimization of the bending energy fosters a proper matching of shape features such as crests, valleys or bumps. In the implementation, a spatial discretization via finite elements, a nonlinear conjugate gradient scheme with a Sobolev metric, and a cascadic multilevel optimization strategy are used. The features of the proposed method are discussed via applications both for synthetic and realistic examples.
Optimising Aperture Shapes for Depth Estimation
(The Eurographics Association, 2013) Sellent, Anita; Favaro, Paolo; Michael Bronstein and Jean Favre and Kai Hormann
The finite depth of field of a real camera can be used to estimate the depth structure of a scene. While the distance of an object from the plane in focus determines the defocus blur size, the shape of the aperture determines the shape of the blur. This blur shape can be manipulated by introducing masks into the main lens aperture. We propose an intuitive criterion to design aperture patterns for depth estimation. Our design criterion imposes constraints directly in the data domain and optimises the amount of depth information carried by blurred images. As a quadratic function on the aperture transmission values, our criterion can be numerically evaluated to estimate optimised aperture patterns quickly. The proposed mask optimisation procedure is applicable for different depth estimation scenarios. We consider depth estimation from two images with different focus settings and depth estimation from two images with different aperture shapes.
Integrated Multi-aspect Visualization of 3D Fluid Flows
(The Eurographics Association, 2013) Brambilla, Andrea; Andreassen, Øyvind; Hauser, Helwig; Michael Bronstein and Jean Favre and Kai Hormann
The motion of a fluid is affected by several intertwined flow aspects. Analyzing one aspect at a time can only yield partial information about the flow behavior. More details can be revealed by studying their interactions. Our approach enables the investigation of these interactions by simultaneously visualizing meaningful flow aspects, such as swirling motion and shear strain. We adopt the notions of relevance and coherency. Relevance identifies locations where a certain flow aspect is deemed particularly important. The related piece of information is visualized by a specific visual entity, placed at the corresponding location. Coherency instead represents the homogeneity of a flow property in a local neighborhood. It is exploited in order to avoid visual redundancy and to reduce occlusion and cluttering. We have applied our approach to three CFD datasets, obtaining meaningful insights.
A Semi-Automated Method for Subject-Specific Modeling of the Spinal Canal from Computed Tomography Images and Dynamic Radiographs
(The Eurographics Association, 2013) Haque, Md. Abedul; Marai, G. Elisabeta; Michael Bronstein and Jean Favre and Kai Hormann
Shrinkage of the spinal canal may be due to congenital or degenerative conditions, and it causes many spinerelated diseases. We present a semi-automated method to computationally reconstruct spinal canal models from static 3D images and dynamic 2D radiographs of the spine. First, we reconstruct the 3D motion of vertebrae from dynamic radiographs and compute hybrid representations of 3D bone models to facilitate computational modeling. We then use the bone position and orientation and the hybrid representations to computationally reconstruct the mesh structure of the spinal canal across the range of motion. The process requires selecting manually only a few landmark points (approximately 1%-2% of all computed vertices), and thus significantly reduces the amount of manual labor required for reconstructing a detailed geometrical model of the spinal canal. Validation on both a healthy and a fusion patient shows that the generated models can capture subject-specific characteristics of the canals and provide insight into the change of the motion pattern due to the surgery. The automation of the method will allow bioengineers to perform large scale experiments on healthy and injured spine joints and thus gain insight into underlying canal conditions.
Mobile Image Retargeting
(The Eurographics Association, 2013) Graf, Daniel; Panozzo, Daniele; Sorkine-Hornung, Olga; Michael Bronstein and Jean Favre and Kai Hormann
We propose an algorithm for axis-aligned content-aware image retargeting that is specifically optimized for mobile devices, and we show that interactive image retargeting is possible even with a low-power, mobile CPU. Our retargeting operator is based on non-uniform scaling and cropping and produces results that are on par with state-of-the-art on a large collection of images. Taking the limited screen space of mobile devices into account, we design a novel user interface that allows painting the saliency map directly onto the retargeted image while the system is continuously recomputing the retargeted result at interactive rates. Finally, we apply our algorithm in a picture gallery application to greatly improve the screen space utilization in mobile device settings.
Datasets and Benchmarks for Densely Sampled 4D Light Fields
(The Eurographics Association, 2013) Wanner, Sven; Meister, Stephan; Goldluecke, Bastian; Michael Bronstein and Jean Favre and Kai Hormann
We present a new benchmark database to compare and evaluate existing and upcoming algorithms which are tailored to light field processing. The data is characterised by a dense sampling of the light fields, which best fits current plenoptic cameras and is a characteristic property not found in current multi-view stereo benchmarks. It allows to treat the disparity space as a continuous space, and enables algorithms based on epipolar plane image analysis without having to refocus first. All datasets provide ground truth depth for at least the center view, while some have additional segmentation data available. Part of the light fields are computer graphics generated, the rest are acquired with a gantry, with ground truth depth established by a previous scanning of the imaged objects using a structured light scanner. In addition, we provide source code for an extensive evaluation of a number of previously published stereo, epipolar plane image analysis and segmentation algorithms on the database.
Epipolar Plane Image Refocusing for Improved Depth Estimation and Occlusion Handling
(The Eurographics Association, 2013) Diebold, Maximilian; Goldluecke, Bastian; Michael Bronstein and Jean Favre and Kai Hormann
In contrast to traditional imaging, the higher dimensionality of a light field offers directional information about the captured intensity. This information can be leveraged to estimate the disparity of 3D points in the captured scene. A recent approach to estimate disparities analyzes the structure tensor and evaluates the orientation on epipolar plane images (EPIs). While the resulting disparity maps are generally satisfying, the allowed disparity range is small and occlusion boundaries can become smeared and noisy. In this paper, we first introduce an approach to extend the total allowed disparity range. This allows for example the investigation of camera setups with a larger baseline, like in the Middlebury 3D light fields. Second, we introduce a method to handle the difficulties arising at boundaries between fore- and background objects to achieve sharper edge transitions.
Fine-Scale Editing of Continuous Volumes using Adaptive Surfaces
(The Eurographics Association, 2013) Ruhl, Kai; Wenger, Stephan; Franke, Dennis; Saretzki, Julius; Magnor, Marcus; Michael Bronstein and Jean Favre and Kai Hormann
Many fields of science such as astronomy and astrophysics require the visualization and editing of smooth, continuous volume data. However, current high-level approaches to volume editing concentrate on segmentable volume data prevalent in medical or engineering contexts, and therefore rely on the presence of well-defined 3D surface layers. Editing arbitrary volumes, on the other hand, is currently only possible using low-level approaches based on the rather unintuitive direct manipulation of axis-aligned slices. In this paper, we present a technique to add or modify fine-scale structures within astronomical nebulae based on adaptive drawing surfaces that enable 2Dimage- like editing approaches. Our results look more natural and have been produced in a much shorter time than previously possible with axis-aligned slice editing.
Cloth-Fluid Contact
(The Eurographics Association, 2013) Huber, Markus; Eberhardt, Bernhard; Weiskopf, Daniel; Michael Bronstein and Jean Favre and Kai Hormann
We present a robust and efficient method for the two-way coupling between particle-based fluid simulations and infinitesimally thin solids represented by triangular meshes. Our approach is based on a hybrid method that combines a repulsion force approach with a continuous intersection handling to guarantee that no penetration occurs. Moreover, boundary conditions for the tangential component of the fluids velocity are implemented to model the no-slip boundary condition. The proposed method is particularly useful for dynamic surfaces, like cloth and thin shells. In addition, we demonstrate how standard fluid surface reconstruction algorithms can be modified to prevent the calculated surface from intersecting close objects. We have implemented our approach for the bidirectional interaction between liquid simulations based on Smoothed Particle Hydrodynamics (SPH) and standard mesh-based cloth simulation systems.
Adaptive Surface Visualization of Vessels with Embedded Blood Flow Based on the Suggestive Contour Measure
(The Eurographics Association, 2013) Lawonn, Kai; Gasteiger, Rocco; Preim, Bernhard; Michael Bronstein and Jean Favre and Kai Hormann
The investigation of hemodynamic information for the assessment of cardiovascular diseases (CVD) has increased in recent years. Improved flow measuring modalities and computational fluid dynamics (CFD) simulations are suitable to provide domain experts with reliable blood flow information. For a visual exploration of the flow information domain experts are used to investigate the flow information combined with its enclosed vessel anatomy. Since the flow is spatially embedded in the surrounding vessel surface, occlusion problems have to be resolved that include a meaningful visual reduction of the vessel surface but still provide important anatomical features. We accomplish this by applying an adaptive surface visualization inspired by the suggestive contour measure. Our approach combines several visualization techniques to improve the perception of surface shape and depth. Thereby, we ensure appropriate visibility of the embedded flow information, which can be depicted with established or advanced flow visualization techniques. We apply our approach to cerebral aneurysms and aortas with simulated and measured blood flow. In an informal user feedback with nine domain experts, we confirm the advantages of our approach compared with existent methods, e.g., semi-transparent surface rendering.
Towards Multi-Kernel Ray Tracing for GPUs
(The Eurographics Association, 2013) Schiffer, Thomas; Fellner, Dieter W.; Michael Bronstein and Jean Favre and Kai Hormann
Ray tracing is a widely used algorithm to compute images with high visual quality. Mapping ray tracing computations to massively parallel hardware architectures in an efficient manner is a difficult task. Based on an analysis of current ray tracing algorithms on GPUs, a new ray traversal scheme called batch tracing is proposed. It decomposes the task into multiple kernels, each of which is designed for efficient execution. Our algorithm achieves comparable performance to state-of-the-art approaches and represents a promising avenue for future research.
A Tracking Approach for the Skeletonization of Tubular Parts of 3D Shapes
(The Eurographics Association, 2013) Garro, Valeria; Giachetti, Andrea; Michael Bronstein and Jean Favre and Kai Hormann
In this paper we propose a new simple and efficient method to characterize shapes by segmenting their elongated parts and characterizing them with their centerlines. We call it Tubular Section Tracking, because it consists of slicing the interested volume along different directions, tracking centroids of the extracted sections with approximately constant centroid position, area and eccentricity and refining the extracted lines with a post processing step removing bad branches and centering, joining and extending the relevant ones. We show that, even using just a few slicing directions (in some cases even just three perpendicular directions), the method is able to obtain good results, approximately pose independent and that the extracted lines can be more informative on the relevant feature of the objects than the classical skeletal lines extracted as subsets of the medial axis. Estimated lines can be used to segment shapes into meaningful parts and compute useful parameters (e.g. length, diameters).

Browse

Browsing VMV13 by Issue Date

Results Per Page

Sort Options