In molecular biology, illustrative animations are used to convey complex biological phenomena to broad audiences. However, such animations have to be manually authored in 3D modeling software, a time consuming task that has to be repeated from scratch for every new data set, and requires a high level of expertise in illustration, animation, and biology. We therefore propose metamorphers : a set of operations for defining animation states as well as the transitions to them in the form of re-usable storytelling templates. The re-usability is two-fold. Firstly, due to their modular nature, metamorphers can be re-used in different combinations to create a wide range of animations. Secondly, due to their abstract nature, metamorphers can be re-used to re-create an intended animation for a wide range of compatible data sets. Metamorphers thereby mask the low-level complexity of explicit animation specifications by exploiting the inherent properties of the molecular data, such as the position, size, and hierarchy level of a semantic data subset. We demonstrate the re-usability of our technique based on the authoring and application of two animation use-cases to three molecular data sets.
Quantum chromodynamics, most commonly referred to as QCD, is a relativistic quantum field theory for the strong interaction between subatomic particles called quarks and gluons. The most systematic way of calculating the strong interactions of QCD is a computational approach known as lattice gauge theory or lattice QCD. Space-time is discretised so that field variables are formulated on the sites and links of a four dimensional hypercubic lattice. This technique enables the gluon field to be represented using 3 × 3 complex matrices in four space-time dimensions. Importance sampling techniques can then be exploited to calculate physics observables as functions of the fields, averaged over a statistically-generated and suitably weighted ensemble of field configurations. In this paper we present a framework developed to visually assist scientists in the analysis of multidimensional properties and emerging phenomena within QCD ensemble simulations. Core to the framework is the use of topology-driven visualisation techniques which enable the user to segment the data into unique objects, calculate properties of individual objects present on the lattice, and validate features detected using statistical measures. The framework enables holistic analysis to validate existing hypothesis against novel visual cues with the intent of supporting and steering scientists in the analysis and decision making process. Use of the framework has lead to new studies into the effect that variation of thermodynamic control parameters has on the topological structure of lattice fields.
Spinel-group minerals are excellent indicators of geological environments and are of invaluable help in the search for mineral deposits of economic interest. The geologists analyze them by means of Barnes and Roeder's contours. In this paper, we present a collection of novel, interactive methods, which assist geologists in the categorization of spinel-group minerals. We fully integrate Barnes and Roeder's contours using a polygonal representation. This makes it possible to efficiently superimpose user-provided point data over the contours, and to automatically rank the contours based on the number of enclosed points. We also allow the expert to create contours for the user-provided point data. Once user contours are created, they can be compared with Barnes and Roeder's contours. During the analysis, the user can drill-down by means of brushing. As we deal with specific data, we apply two novel brushing techniques, i.e., the percentile brush and the contour brush. The novel brushing mechanisms along with the interactive comparison speed-up the analysis significantly. We evaluate the newly introduced approach and the resulting novel workflow using real-word data from different locations in Argentina. According to the domain experts, the classification of spinel minerals needs several minutes now, while it took a few days with the current state of the art approach in the domain.
Visualizations make rich use of multiple visual channels so that there are few resources left to make selected focus elements visually distinct from their surrounding context. A large variety of highlighting techniques for visualizations has been presented in the past, but there has been little systematic evaluation of the design space of highlighting. We explore highlighting from the perspective of visual marks and channels - the basic building blocks of visualizations that are directly controlled by visualization designers. We present the results from two experiments, exploring the visual prominence of highlighted marks in scatterplots: First, using luminance as a single highlight channel, we found that visual prominence is mainly determined by the luminance difference between the focus mark and the brightest context mark. The brightness differences between context marks and the overall brightness level have negligible influence. Second, multi-channel highlighting using luminance and blur leads to a good trade-off between highlight effectiveness and aesthetics. From the results, we derive a simple highlight model to balance highlighting across multiple visual channels and focus and context marks, respectively.
Histological serial sections allow for 3D representation of anatomical structures in microscopic to mesoscopic range. However, due to the nature of the acquisition, they suffer from severe anisotropy: 14-to-1 in a single average microscopic paraffin section. We present an interpolation method based on optical flow and show that standard interpolation methods are less suited for serial sections. With our non-linear interpolation approach we are able to represent the "movement" of image parts that are of interest. This allows for better 3D reconstructions and further insights in microanatomy.
Multiple-run simulations are widely used for investigation of dynamic systems, where they combine varied input parameters and different kinds of outputs. In this work, we focus on a simulation type that outputs an ensemble of surfaces for each simulation run. Multiple simulation runs, in this case, result in a set of surface ensembles - a super-ensemble. We propose an advanced data model, abstract analysis tasks, and introduce an analysis workflow for the exploration of super-ensembles. To address the challenging exploration and analysis tasks, we present Super-Ensembler, a visual analytics system for analysis of data-surface collections as super-ensembles. We introduce novel aggregation methods and corresponding visualizations. The aggregation techniques reduce data complexity by either yielding a super-ensemble of a simplified data type or a conventional surface ensemble. Novel visual representations include an overview visualization for super-ensembles, 3D multi-resolution box plots, and intersection contours. Together with standard views, such as scatter plots, parallel coordinates, or histograms, they are integrated into a coordinated multiple views framework. The newly proposed methodology is developed in a close collaboration with experts from the automotive domain. We evaluate our approach by means of a case study in the context of gear transmission design. Positive feedback and reported speed-up of the analysis indicate the usefulness of the presented approach.
Digital approaches to shape comparison and analysis play a very important role in forensic anthropology. New methods are still emerging and the whole area is experiencing a shift from traditional 2D image data to processing of 3D meshes. Therefore, the visual exploration of 3D meshes and methods for their visual comparison play a crucial role in the anthropological research. In our paper we present a novel AnthroVis tool for visual analysis of 3D mesh ensembles, which was designed in tight cooperation with the domain experts. It aims to enhance their workflow by introducing several visualizations that help to understand the similarities and differences between 3D meshes. AnthroVis in general consists of three methods, which serve as a guidance in the process of the comparison of two or more mesh ensembles. The first method, based on the idea of interactive heat plots, provides an overview of pairwise comparisons in a set of analyzed meshes and enables their filtering and sorting. The second method consists of anthropologically relevant cross-cuts indicating the variability through the set of meshes. The last method uses superimposition principle for pairs of meshes equipped with several visual enhancements indicating local mesh differences in three-dimensional space. The domain expert evaluation was performed primarily on facial images, but the tool proved to be applicable to other areas of forensic anthropology as well. Its usefulness is demonstrated by three case studies describing the real situations and problems encountered by anthropologists in forensic casework.
HDR video acquisition and processing are quite challenging tasks from the point of both computational complexity and algorithm design. In this paper, we present a real-time HDR processing system evaluated on a custom hardware camera platform. We exploited well known state of the art algorithms for HDR acquisition from sequence of multiexposure frames and real-time tone-mapping. We also present modifications of the algorithms enabling efficient implementation on FPGA platform and real-time performance. The main focus of the paper is on acceleration of Durand local tone-mapping operator involving real-time bilateral filter. The proposed solution is compared to existing research results in terms of speed, resource consumptions, and numerical accuracy.
In this work, we investigate the application of the Bag-of-Words approach for object search task in 3D domain. Image retrieval task solutions, operating on datasets of thousands and millions images, have proved the effectiveness of Bag-of-Words approach. The availability of low cost RGB-D cameras is a rise of large datasets of 3D data similar to image corpuses (e.g. RoboEarth). The results of such an investigation could be useful for many robot scenarios like place recognition from a large dataset of samples of places acquired during the long-term observation of an environment. The first goal of our research presented in this paper is focused on the sensitivity of the Bag-of-Words approach to various parameters (e.g. spacial sampling, surface description etc.) with respect to precision, stability and robustness. The experiments are carry out on two widely-used datasets in object instance identification task in 3D domain.
This paper proposes an agent-based model for animating molecular machines. Usually molecular machines are visualized using key-frame animation. Creating large molecular assemblies with key-frame animation in standard 3D software can be a tedious task, because hundreds or thousands of molecular particles have to be animated by hand, considering various biological phenomena. To avoid repetitive animation of molecular particles, a prototypic framework is implemented, that employs an agent-based approach. Instead of animating the molecular particles directly, the framework utilizes a behavior description for each type of molecular particle. The animation results from the molecular particles interacting with each other as defined by their behavior. Interaction between molecular particles is enabled by an abstract model that is implemented by the framework. The methodology for creating the framework was driven through learning by example. Three molecular machines are visualized using the framework. During this process, the framework was iteratively improved and extended. The resulted animations demonstrate that agent-based animation is a viable option for molecular machines.
Visually accurate capture of appearance of highly specular surfaces is of a great research interest of the coating industry, who strive to introduce highly reflective products while minimizing their production and quality assessment costs, and avoiding environmental issues related to the production process. An efficient measurement of such surfaces is challenging due to their narrow specular peak of an unknown shape and typically very high dynamic range. Such behavior puts higher requirements on capabilities of a measuring device and has impact on length of the measurement process. In this paper, we rely on a material probes with a predefined curved shape featuring slight local inhomogeneities. This defines a goniometric device as appropriate means of appearance capture. To shorten a typically long measurement time required when using these approaches, we introduce a method of material appearance acquisition by means of the isotropic BRDF using relatively sparse sampling adapted to each measured material individually.
We present a system that automatically suggests the furniture layout when one moves into a new house, taking into account the furniture layout in the previous house. In our method, the input to our system comprises the floor plans of the previous and new houses, and the furniture layout in the previous house. The furniture layout for the whole house is suggested. This method builds on a previous furniture layout method with which the furniture layout for a single room only is computed. In this paper, we propose a new method that can suggest the furniture layout for multiple rooms in the new house. To deal with this problem, we took a heuristic approach in developing a cost function by adding some new cost functions to the previous method. We show various layouts computed using our method, which demonstrates the effectiveness of it.
In this paper we present an automated approach for optimizing the conspicuity of features in 3D volume visualization. By iteratively adjusting the opacity transfer function, we are able to generate visualizations that satisfy a user-specified target distribution defining the relative conspicuity of particular features in the data set. Our approach exploits a metric, called Visibility-Weighted Saliency (VWS), that takes into account both the issues of view-dependent occlusion and visual saliency in defining the visibility of features in volume data. A parallel line search strategy is presented to improve the performance of the optimization mechanism. We demonstrate that the approach is able to achieve promising results in optimizing visualizations of both static and time-varying volume data.
In Virtual Reality (VR), the action of selecting virtual objects outside arms-reach still poses significant challenges. In this work, after classifying, with a new taxonomy, and analyzing existing solutions, we propose a novel technique to perform out-of-reach selections in VR. It uses natural pointing gestures, a modifiable cone as selection volume, and an iterative progressive refinement strategy. This can be considered a VR implementation of a discrete zoom approach, although we modify users' position instead of the field-of-view. When the cone intersects several objects, users can either activate the refinement process, or trigger a multiple object selection. We compared our technique against two techniques from literature. Our results show that, although not being the fastest, it is a versatile approach due to the lack of errors and uniform completion times.
Cage-based structures are reduced subspace deformers enabling non-isometric stretching deformations induced by clothing or muscle bulging. In this paper, we reformulate the cage-based rigging as an incompressible Stokes problem in the vorticity space. The key to our approach is a compact stencil allowing the expression of fluid-inspired high-order coordinates. Thus, our cage-based coordinates are obtained by vorticity transport as the numerical solution of the linearized Stokes equations. Then, we turn the incompressible creeping Newtonian flow into Stokes equations, and we devise a second-order compact approximation with center differencing for solving the vorticity-stream function. To the best of our knowledge, our work is the first to devise a vorticity-stream function formulation as a computational model for cage-based weighting functions. Finally, we demonstrate the effectiveness of our new techniques for a collection of cage-based shapes and applications.
Using word clouds to visualize dynamic time-varying data is a field still under-explored. The goal of our approach is to provide a novel way of generating smoothly animated word clouds to show changes in word frequency via font size. Unlike existing methods, a compact layout, inspired by the popular word cloud generation tool Wordle, is preserved during animation and implemented using web technologies. Word size changes in time are also illustrated via color and word rotation.
In visualization systems it is often the case that the changes of the input parameters are not proportional to the visual change of the generated output. In this paper, we propose a model for enabling data-sensitive navigation for user-interface elements. This model is applied to normalize the user input according to the visual change, and also to visually communicate this normalization. In this way, the exploration of heterogeneous data using common interaction elements can be performed in an efficient way. We apply our model to the field of medical visualization and present guided navigation tools for traversing vascular structures and for camera rotation around 3D volumes. The presented examples demonstrate that the model scales to user-interface elements where multiple parameters are set simultaneously.
The realistic simulation of clouds in synthetic environments has always been an important research topic in computer graphics. However, simulation of clouds on computer is a complicated task since they have physics-dependent dynamic shapes that evolve over time. Cloud simulation involves two distinct fields: computer graphics and meteorology. While the first focuses on representing clouds in a visually appealing way, the second focuses on the physics of clouds and weather forecasts. Both perspectives are integrated in our cloud simulator. It takes advantage of SkewT/LogP (Skewed Temperature and Logarithmic Pressure) diagrams [Air Weather Service HQ 1961], also known as thermodynamic diagrams, from which we take temperature and pressure values as inputs to feed the cloud motion equation -without the need for solving partial differential equations- that regulates the flow of clouds as fluids in the atmosphere, as well as to generate cloud shapes and dynamics that can be easily integrated in flight simulators or computer games. As illustrated in Fig. 1, the main contributions of our work are as follows: • SkewT/LogP diagrams. To our best knowledge, this is the first cloud simulator that uses SkewT/LogP diagrams in computer graphics. In fact, we have built a visual tool for 2D SkewT/LogP diagrams that allows us to inspect, control and simulate the thermodynamic process of ascending clouds in the atmosphere, as well as a 3D synthetic environment where clouds are advected by buoyant forces. • Solving the equation of motion for clouds in real-time. By using SkewT/LogP diagrams, we are able to explicitly determine the vertical rising force required to solve the equation of motion of an air parcel (mass of air) in the atmosphere, without solving its differential equations. • Automated generation of clouds that adapt to weather conditions. By using real weather data (including wind data) mapped onto SkewT/LogP diagrams, we are able to generate realistic synthetic clouds from the ground to their equilibrium level (EL) in the atmosphere.
We introduce geometric graph grammars, demonstrate how they can generate geometric structures, and introduce an algorithm for their automatic learning (inverse procedural modeling). Our approach extends the concept of graph grammars to allow for coding not only topological data, but also geometry. Forward modeling generates geometric graphs and considers various strategies for node connectivity. Inverse procedural modeling performs learning of geometric graphs, by discovering repeated structures and their connectivity. These structures are encoded into geometric graph grammar rewriting rules. We demonstrate usability of our approach on an example using urban networks. Graph learning is reasonably fast; in our implementation, learning of a road network with 72k vertices and 100k edges is performed in less than one minute.
This paper tackles the unsolved important problem of training deep models with small amounts of annotated data. We propose a semi-supervised self-training bootstrap to deep learning on small datasets by retrieving and utilizing additional images from internet image search. We adapt the pseudolabel method proposed by Dong-Hyun Lee in 2013, previously used on the elementary MNIST handwritten digit classification task. We show that by suitable modifications to its example weighting and selection mechanisms it can be adapted to general image classification tasks supported by online image search. The proposed approach does not require any human supervision, it is practical and efficient, and it actively avoids overtraining. The usefulness of the proposed method is demonstrated on the SUN 397 dataset with only 50 training images per category. When exploiting results of Google's Image Search, we achieve a significant improvement, with a classification accuracy of 51%, as opposed to 39% without these results.
