WSCG 2017: Full Papers Proceedings

Permanent URI for this collection

http://hdl.handle.net/11025/26391

Browse

Showing 1 - 19 out of 19 results

Improvement of some interpolation methods for terrain reconstruction from scattered data
(Václav Skala - UNION Agency, 2017) Bohdal, Róbert; Skala, Václav
Using GPS modules it is easy to obtain 3D data for areas that have not been digitized yet. Such terrain data are usually not arranged in a grid, and therefore we have to use scattered data interpolation methods. The aim of the paper is to create a digital terrain model from 3D data using modifications of known methods. Sibson interpolation method is often used when we need to interpolate large data sets. This method has low memory requirements, it is sufficiently fast, but creates undesired surface artefacts. Our aim is to have the resulting interpolation surface as similar as possible to the original surface. We have decided to replace the heights at specified points by local functions. We use biquadratic and bicubic polynomials, Hardy’s multiquadrics and thin plate spline as local functions. In the paper, we have evaluated the time requirements and the accuracy with which the interpolated area matches the actual 3D data on 2 terrain samples (the Little Carpathians and a small part of the Little Carpathians).
Pushpins for edit propagation
(Václav Skala - UNION Agency, 2017) Mylo, Marlon; Klein, Reinhard; Skala, Václav
In this paper we present an approach for stroke-input based foreground estimation of measured materials with a near regular structure. To enable extraction of high-quality editing masks even from difficult materials, we combine a state of the art lattice-detection algorithm with a novel frequency convolution scheme, which we call pushpins. Despite being highly specialized, we consider this use-case as important for material design. A comparison with other state of the art editing and material recognition approaches will give proof of the robustness and ability of our algorithm.
StreetGAN: towards road network synthesis with generative adversarial networks
(Václav Skala - UNION Agency, 2017) Hartmann, Stefan; Weinmann, Michael; Wessel, Raoul; Klein, Reinhard; Skala, Václav
We propose a novel example-based approach for road network synthesis relying on Generative Adversarial Networks (GANs), a recently introduced deep learning technique. In a pre-processing step, we first convert a given representation of a road network patch into a binary image where pixel intensities encode the presence or absence of streets. We then train a GAN that is able to automatically synthesize a multitude of arbitrary sized street networks that faithfully reproduce the style of the original patch. In a post-processing step, we extract a graph-based representation from the generated images. In contrast to other methods, our approach does neither require domainspecific expert knowledge, nor is it restricted to a limited number of street network templates. We demonstrate the general feasibility of our approach by synthesizing street networks of largely varying style and evaluate the results in terms of visual similarity as well as statistical similarity based on road network similarity measures.
Fast and effective dynamic mesh completion
(Václav Skala - UNION Agency, 2017) Arvanitis, Gerasimos; Lalos, Aris S.; Moustakas, Konstantinos; Fakotakis, Nikos; Skala, Václav
We introduce a novel approach to support fast and efficient completion of arbitrary animation sequences, ideally suited for real-time scenarios, such as immersive tele-presence systems and gaming. In most of these applications, the reconstruction of 3D animations is based on dynamic meshes which are highly incomplete, stressing the need of completion approaches with low computational requirements. In this paper, we present a new online approach for fast and effective completion of 3D animated models that estimates the position of the unknown vertices of the current frame by exploiting the connectivity information and the current motion vectors of the known vertices. Extensive evaluation studies carried out using a collection of different incomplete animated models, verify that the proposed technique achieves plausible reconstruction output despite the constraints posed by arbitrarily complex and motion scenarios.
iDotter: an interactive dot plot viewer
(Václav Skala - UNION Agency, 2017) Gerighausen, Daniel; Hausdorf, Alrik; Zänker, Sebastian; Skala, Václav
Bioinformaticians judge the likelihood of the overall RNA secondary structure based on comparing its base pair probabilities. These probabilities can be calculated by various tools and are frequently displayed using dot plots for further analysis. However, most tools produce only static dot plot images which restricts possible interactions to the capabilities of the respective viewers (mostly PostScript-viewers). Moreover, this approach does not scale well with larger RNAs since most PostScript viewers are not designed to show a huge number of elements and have only legacy support for PostScript. Therefore, we developed iDotter, an interactive tool for analyzing RNA secondary structures. iDotter overcomes the previously described limitations providing multiple interaction mechanisms facilitating the interactive analysis of the displayed data. According to the biologists and bioinformaticians that regularly use out interactive dot plot viewer, iDotter is superior to all previous approaches with respect to facilitating dot plot based analysis of RNA secondary structures.
Accurate triangular regular network adjustment to large digital elevation models
(Václav Skala - UNION Agency, 2017) Santana, José M.; Trujillo, Augustín; Suárez, José P.; Ortega, Sebastián; Skala, Václav
Nowadays, large volumes of terrain data are available to use as Digital Elevation Models, from which coarser meshes can be progressively generated for visualization and other purposes. Previous studies compared different methods to adjust those meshes, concluding that no method performs the best for all kinds of terrain. In this work, a pipeline to accurately adjust TRNs to DEM is proposed. An initial approximation is calculated by solving a linear system from input data. Vertices with a major contribution to the global error of the mesh are then tuned using a local refinement algorithm. Experimentation shows that meshes adjusted using the proposed pipeline fit better the original DEM than ones generated using classic methods as linear interpolation for several benchmark elevation models.
Accelerating radiosity on GPUs
(Václav Skala - UNION Agency, 2017) Shcherbakov, Alexandr; Frolov, Vladimir; Skala, Václav
We propose a novel approach to implement radiosity on GPU with specific optimizations via form-factor matrix transformations. The proposed transformations enable to reduce the amount of computations for multiple-bounce global illumination and apply DXT compression (with subsequent hardware decompression when reading formfactors on GPU). Our implementation is 10 times faster running and requires 3 times less memory than the naive radiosity GPU implementation.
Head movement based temporal antialiasing for VR HMDs
(Václav Skala - UNION Agency, 2017) Kim, Jung-Bum; Choi, Soo-Ryum; Choi, Joon-Hyun; Ahn, Sang-Jun; Park, Chan-Min; Skala, Václav
Inherent properties of VR HMDs cause degradation of visual quality which disrupts immersive VR experience. We identify a new temporal aliasing problem caused by unintended tiny head movement of VR HMD users. The images that users see slightly change, even in the case that the users intend to hold and concentrate on a certain part of VR content. The slight change is more perceivable, because the images are magnified by lenses of VR HMDs. We propose the head movement based temporal antialiasing approach which blends colors that users see in the middle of head movement. In our approach, the way to determine locations and weights of colors to be blended is based on head movement and time stamp. Speed of head movement also determines proportions of colors in the past and at present in blending. The experimental result shows that our approach is effective to reduce the temporal aliasing caused by unintended head movement in real-time performance.
Marine snow detection and removal: underwater image restoration using background modeling
(Václav Skala - UNION Agency, 2017) Farhadifard, Fahimeh; Radolko, Martin; Freiherr von Lukas, Uwe; Skala, Václav
It is a common problem that images captured underwater (UW) are corrupted by noise. This is due to the light absorption and scattering by the marine environment; therefore, the visibility distance is limited up to few meters. Despite blur, haze, low contrast, non-uniform lightening and color cast which occasionally are termed noise, additive noises, such as sensor noise, are the center of attention of denoising algorithms. However, visibility of UW scenes is distorted by another source termed marine snow. This signal not only distorts the scene visibility by its presence but also disturbs the performance of advanced image processing algorithms such as segmentation, classification or detection. In this article, we propose a new method that removes marine snow from successive frames of videos recorded UW. This method utilizes the characteristics of such a phenomenon and detects it in each frame. In the meanwhile, using a background modeling algorithm, a reference image is obtained. Employing this image as a training data, we learn some prior information of the scene and finally, using these priors together with an inpainting algorithm, marine snow is eliminated by restoring the scene behind the particles.
Speeding up the computation of uniform bicubic spline surfaces
(Václav Skala - UNION Agency, 2017) Kačala, Viliam; Miňo, Lukáš; Skala, Václav
Approximation of surfaces plays a key role in a wide variety of computer science fields such as graphics or CAD applications. Recently a new algorithm for evaluation of interpolating spline surfaces with C2 continuity over uniform grids was proposed based on a special approximation property between biquartic and bicubic polynomials. The algorithm breaks down the classical de Boor’s computational task to reduced tasks and simple remainder ones. The paper improves the reduced part’s implementation, proposes an asymptotic equation to compute the theoretical speedup of the whole algorithm and provides results of computational experiments. Both de Boor’s and our reduced tasks involves tridiagonal linear systems. First of all, a memory-saving optimization is proposed for the solution of such equation systems. After setting the computational time complexity of arithmetic operations and clarifying the influence of modern microprocessors design on the algorithm’s remainder tasks, a new expression is suggested for assessing theoretical speedup of the whole algorithm. Validity of the equation is then confirmed by measured speedup on various microprocessors.
Predicting vehicle trajectories from surveillance video in a real scenario with Histogram of Oriented Gradient
(Václav Skala - UNION Agency, 2017) Ferreira, Arthur Emidio T.; Espinoza, Bruno Luiggi M.; Barros Vidal, Flavio de; Skala, Václav
We propose a method capable to predict vehicle trajectories in a real scenario based on an unsupervised approach using Histogram of Oriented Gradients (HOG) features to construct an uniform path. The proposed algorithm extracts a sub-region of the input image defined as Field of View of the target vehicle, to output a possible trajectory that the given vehicle will follow through. We perform many experiments using the proposed technique, and based on qualitative/quantitative analyses, we conclude it is successfully able to predict reasonable trajectories.
A primary morphological classifier for skin lesion images
(Václav Skala - UNION Agency, 2017) Macatangay, Jules Matthew A.; Ruiz Jr., Conrado R.; Usatine, Richard P.; Skala, Václav
Classifying skin lesions, abnormal changes in skin, into their morphologies is the first step in diagnosing skin diseases. In dermatology, morphology is a categorization of a skin lesion’s structure and appearance. Rather than directly classifying skin diseases, this research aims to explore classifying skin lesion images into primary morphologies. For preprocessing, k-means clustering for image segmentation and illumination equalization were applied. Additionally, features utilized considered color, texture, and shape. For classification, k-Nearest Neighbors, Decision Trees, Multilayer Perceptron, and Support Vector Machines were used. To evaluate the prototype, 10-fold cross validation was applied over a dataset assembled from online resources. In experimentation, the morphologies considered were macule, nodule, papule, and plaque. Moreover, different feature subsets were tested through feature selection experiments. Experimental results on the 4-class and 3-class tests show that of the classifiers selected, Decision Trees were best, having a Cohen’s kappa of 0.503 and 0.558 respectively.
A generic approach for sunlight and shadow impact computation on large city models
(Václav Skala - UNION Agency, 2017) Jaillot, Vincent; Pedrinis, Frédéric; Servigne, Sylvie; Gesquière, Gilles; Skala, Václav
Study of sunlight and shadow effects on the city has become more accessible with the development of 3D city models. It allows measuring when and how an object is exposed to the sunlight, which enables conducting many related studies such as energy analyses or urban planning. While many works have been done for this purpose, it may be interesting to know which objects (terrain, buildings, trees, etc.) prevent other objects from being exposed to the sunlight. In this paper we propose a method which detects the sunlit zones on a city model and the shadow impact of its objects. As these objects can be of various natures and as the acquisition processes varies from one city to another, they are not all necessarily available in each city model. Since an object’s shadow can impact other very distant objects, we must have a method that handles efficiently large areas, especially knowing that city models can have fine geometric and semantic definitions. The generic approach we propose can manage these different city models by supporting every type of the above-mentioned objects and by relying on the use of standards. This paper presents a generic method which allows sunlight and shadow computation on arbitrarily large 3D city models for impact analyses of each city object on its surroundings (close and far). This means that besides checking if a city object is shaded or not, we know which objects are responsible for the shade, thus allowing various impact analyses on cities.
A novel force feedback haptics system with applications in phobia treatment
(Václav Skala - UNION Agency, 2017) Brice, Daniel; Devine, Scott; Rafferty, Karen; Skala, Václav
It is well known that multi-sensory stimulation can enhance immersion within virtual environments. Whilst there has been rapid development of devices which can enhance the visual immersion, technology to stimulate other senses, such as touch, is still under developed. Currently there is a problem wherein a surface in a virtual environment, such as a wall, cannot replicate the physical properties of a solid object. In this paper a novel system is proposed utilising the HTC VIVE and Rethink Robotics’ Baxter Robot to replicate surfaces. A demonstration has been created whereby a user climbs a wall in a virtual environment by grabbing onto ledges which exist as a physical body located on Baxter’s end effector. The system uses bi-directional TCP communication between an environment developed in Epic Games’ Unreal Engine and the Baxter robot running the Robot Operating System framework. When an ascending user reaches out and grabs a ledge on the virtual wall they will be applying a torque to the Baxter arm which can be measured and the intended movement of the user inferred, resulting in the ledge being moved through a suitable Inverse Kinematics path. This has provided the user with the ability to climb a wall in VR in the absence of any hand tracking methods whilst receiving force feedback from the ledges they grasp onto. Current alternative systems only exist as wearables or operate in small spaces. The increased immersion in this VR demo can be used to assist those with phobias of heights.
A framework for visually realistic multi-robot simulation in natural environment
(Václav Skala - UNION Agency, 2017) Ganoni, Ori; Mukundan, Ramakrishnan; Skala, Václav
This paper presents a generalized framework for the simulation of multiple robots and drones in highly realistic models of natural environments. The proposed simulation architecture uses the Unreal Engine4 for generating both optical and depth sensor outputs from any position and orientation within the environment and provides several key domain specific simulation capabilities. Various components and functionalities of the system have been discussed in detail. The simulation engine also allows users to test and validate a wide range of computer vision algorithms involving different drone configurations under many types of environmental effects such as wind gusts. The paper demonstrates the effectiveness of the system by giving experimental results for a test scenario where one drone tracks the simulated motion of another in a complex natural environment.
Performance analysis of corner detection algorithms based on edge detectors
(Václav Skala - UNION Agency, 2017) Afrin, Naurin; Lai, Wei; Mohammed, Nabeel; Skala, Václav
Detecting corner locations in images plays a significant role in several computer vision applications. Among the different approaches to corner detection, contour-based techniques are specifically interesting as they rely on edges detected from an image, and for such corner detectors, edge detection is the first step. Almost all the contour-based corner detectors proposed in the last few years use the Canny edge detector. There is no comparative study that explores the effect of using different edge detection method on the performance of these corner detectors. This paper fills that gap by carrying out a performance analysis of different contour-based corner detectors when using different edge detectors. We studied four recently developed corner detectors, which are considered as current state of the art and found that the Canny edge detector should not be taken as a default choice and in fact the choice of edge detector can have a profound effect on the corner detection performance. We examined commonly used predefined threshold-based Canny detector with the adaptive Canny detector and found that adaptive Canny detector gives better results to work with.
Flexible moment invariant bases for 2D scalar and vector fields
(Václav Skala - UNION Agency, 2017) Bujack, Roxana; Flusser, Jan; Skala, Václav
Complex moments have been successfully applied to pattern detection tasks in two-dimensional real, complex, and vector valued functions. In this paper, we review the different bases of rotational moment invariants based on the generator approach with complex monomials. We analyze their properties with respect to independence, completeness, and existence and present superior bases that are optimal with respect to all three criteria for both scalar and vector fields.
Increasing diversity and usability of crowd animation systems
(Václav Skala - UNION Agency, 2017) Paduraru, Ciprian; Skala, Václav
Crowd systems are a vital part in virtual environment applications that are used in entertainment, education, training or different simulation systems such as evacuation planning. Because performance and scalability are key factors, the implementation of crowds poses many challenges in many of its aspects: behaviour simulation, animation, and rendering. This paper is focusing on a different model of animating crowd characters that support dynamically streaming of animation data between CPU and GPU. There are three main aspects that motivate this work. First, we want to provide a greater diversity of animations for crowd agents than was possible before, by not storing any predefined animation data. Another aspect that stems from the first improvement is that this new model allows the crowd simulation side to communicate more efficiently with the animation side by sending events back and forth at runtime, fulfilling this way use-cases that different crowd systems have. Finally, a novel technique implementation that blends between crowd agents’ animations is presented. The results section shows that these improvements are added with negligible cost.
Image-based information visualization: (or how to unify SciVis and InfoVis)
(Václav Skala - UNION Agency, 2017) Telea, Alexandru; Skala, Václav
For decades, scientific visualization (SciVis) and information visualization (InfoVis) have been related, but still distinctly separated disciplines. Methods and techniques in the two areas have developed relatively separately, causing an arguably unnecessarily separation in the visualization field. Attempts for unification exist, but are largely based on heuristics, and subject to critique from both the SciVis and InfoVis angles. In this talk, we argue that this separation is not necessary, and, up to large extents, artificial. More specifically, we argue that the difference between SciVis and InfoVis is not a matter of design decisions only, but, more centrally, a matter of representing the structure of large data collections by means of smooth, continuous, encodings. We present a way to cast InfoVis along the same principles as the more classical SciVis, based on a continuous, multiscale, spatial representation of data. Putting it simply, we argue that visualizing large amounts of InfoVis data can use encoding techniques which share the same continuity and multiscale principles as most classical spatial SciVis (or image processing) methods use. In turn, we show how this is possible by means of defining appropriate similarity metrics and encoding principles for InfoVis data. This leverages a wealth of data simplification, encoding, and perception principles, since long available for SciVis data, for the richer realm of InfoVis data. We demonstrate our imagebased paradigm by examples covering the visualization of relational, multidimensional, and time-dependent InfoVis.

Browse

Recent Submissions