WSCG 2015: Full Papers Proceedings

Permanent URI for this collection

http://hdl.handle.net/11025/26387

Browse

Showing 1 - 20 out of 29 results

Vision and virtual-based human computer interaction applications for a new digital media visualization
(Václav Skala - UNION Agency, 2015) Kerdvibulvech, Chutisant; Skala, Václav; Gavrilova, Marina
With the rise of smartphones and tablets interactively, human computer interaction is a very popular topic for engineers, artists, designers and computer scientists around the world in both industry and academia. This topic was studied and researched over many years ago. Nevertheless, most of previous works were studied separately between communication arts (e.g., advertising and marketing communication research) and computer science. Indeed, there has been little work giving an overview of recent integrated research of digital media and some new technologies, such as computer vision, virtual reality, and human computer interaction for visual communication. Therefore, our contribution of this paper is to discuss the recent state-of-the-art development of the digital media research work using and applying these aforementioned multimedia-based technologies. A literature review of the novel digital media and interactive augmented reality researches is also discussed. More importantly, this paper also provides a work-in-progress framework for future digital media research when applying graphical visualization, human computer interaction such as haptic, and sensor technologies into every traditional sense of human interactively, from vision to touch and from smell to taste. In general, this paper will be beneficial for any related field of interactive multimedia, communication arts and human computer interaction both industrial and educational aspects and also for any related researcher such as computer science art communicator.
Online 3D signature verification by using stereo camera & tablet
(Václav Skala - UNION Agency, 2015) Dave, Jay; Venkatesh, KS; Jain, Garima; Skala, Václav; Gavrilova, Marina
The signature of a person is an important biometric attribute which can be used to authenticate human identity. Conventional online approaches to signature verification only use either a single camera to track the pen tip position or a tablet to extract the dynamic features of the signature, hence the signature has only two spatial dimensions. In this paper we combine data inputs from a pressure sensitive device (tablet digitizer) and stereo vision to record signatures in 3D. Stereo vision from a pair of low cost SONY Eyecam cameras is used to track the pen tip position in x, y, & in z when the pen is off the surface as well as the pen angle with respect to the surface at all times. The digitizing tablet on the other hand, tracks x, y as well as pressure magnitude (which we denote as 􀀀z) when the pen contacts the surface. In all, we record the following parameters as functions of time through the duration of the signature: x;y; z;q;f, where all the linear paramaters are bipolar, with the particular case of z representing motion with positive values and pressure level with negative values. The angular values are two dimensional. The distance between the input signature’s features recorded as a 5-variate parameter time sequence and the template signature’s features whichwere collected during the training phase is computed using Dynamic Time Warping (DTW), and is thresholded to take a decision. While better learning techniques and more intensive experimentation will help suggest improvements, even as of the present, we have a fully working prototype of the system.
Hybrid client-server and P2P network for web-based collaborative 3D design
(Václav Skala - UNION Agency, 2015) Desprat, Caroline; Luga, Hervé; Jessel, Jean-Pierre; Skala, Václav; Gavrilova, Marina
Our proposed research project is to enable 3D distributed visualization and manipulation involving collaborative effort through the use of web-based technologies. Our project resulted from a wide collaborative application research fields: Computer Aided Design (CAD), Building Information Modeling (BIM) or Product Life Cycle Management (PLM) where design tasks are often performed in teams and need a fluent communication system. The system allows distributed remote assembling in 3D scenes with real-time updates for the users. This paper covers this feature using hybrid networking solution: a client-server architecture (REST) for 3D rendering (WebGL) and data persistence (NoSQL) associated to an automatically built peer-to-peer (P2P) mesh for real-time communication between the clients (WebRTC). The approach is demonstrated through the development of a web-platform prototype focusing on the easy manipulation, fine rendering and light update messages for all participating users. We provide an architecture and a prototype to enable users to design in 3D together in real time with the benefits of web based online collaboration.
Adaptive depth bias for soft shadows
(Václav Skala - UNION Agency, 2015) Ehm, Alexander; Ederer, Alexander; Klein, Andreas; Nischwitz, Alfred; Skala, Václav; Gavrilova, Marina
With shadow mapping the need of a suitable biasing technique due to shadow aliasing is indisputable. Dou et al. [Dou14] introduced a biasing technique that computes the optimal bias adaptively for each fragment. In this paper, we propose enhancements for this algorithm. First, we extend the algorithm for soft shadows, such as percentage closer filtering (PCF) and percentage closer soft shadows (PCSS). Second, we minimize the projective aliasing by introducing a scale factor depending on the ratio between surface and light direction. We show that our enhancements increase the shadow quality and introduce only a small overhead.
An efficient reduction of IMU drift for registration error free augmented reality maintenance application
(Václav Skala - UNION Agency, 2015) Lakshmiprabha, N. S.; Santos, Alexander; Beltramello, Olga; Skala, Václav; Gavrilova, Marina
Augmented reality (AR) is a technology that overlays virtual 3D content in the real world to enhance a user’s perception. This AR virtual content must be registered properly with less jitter, drift or lag to create a more immersive feeling for the user. The object pose can be determined using different pose estimation techniques using the data from sensors cameras and inertial measurement units (IMUs). Camera based vision algorithms detect the features in a given environment to calculate the relative pose of an object with respect to the camera. However, these algorithms often take a longer time to calculate the pose and can only operate at lower rates. On the other hand, an IMU can provide fast data rates from which an absolute pose can be determined with fewer calculations. This pose is usually subjected to drift which leads to registration errors. The IMU drift can be substantially reduced by fusing periodic pose updates from a vision algorithm. This work investigates various factors that affect the rendering registration error and to find the trade-off between the vision algorithm pose update rate and the IMU drift to efficiently reduce this registration error. The experimental evaluation details the impact of IMU drift with different vision algorithm pose update rates. The results show that the careful selection of vision algorithm pose updates not only reduces IMU drift but also reduces the registration error. Furthermore, this reduces the computation required for processing the vision algorithm.
Corrosion rendering: fusing simulation and photo-texturing
(Václav Skala - UNION Agency, 2015) Jain, Nisha; Kalra, Prem; Subodh, Kumar; Skala, Václav; Gavrilova, Marina
We present a technique for realistic rendering of corroded objects. We employ a physio-chemically based stochastic model to determine the deterioration level of different points on an object, given its material characteristics and the vigor of the environment. Guided by values from the ISO standard, our model predicts shape degradation. This shape degradation is then applied to the object in the form of surface displacements and weathered appearance. The appearance degradation is hard to physically model accurately due to its dependence on a large number of unknown parameters as well as its high sensitivity to errors in modeling them. Hence, we instead sample from photographs of real objects to generate similar appearance for the rendered surface, but consistent with the simulated corrosion levels. We demonstrate our technique using several simulation results as well as different input photographs. We also evaluate the fidelity of the generated output to the simulation as well as to the sample texture patterns and validate our work with the help of data published in the corrosion literature. Our framework is generic and can be extended to a variety of corrosion scenarios. Ours is an important step towards predictive analysis of material loss and weathering phenomena for real objects.
Robust temporal depth enhancement method for dynamic virtual view synthesis
(Václav Skala - UNION Agency, 2015) Liu, Can; Zhang, Weizheng; Skala, Václav; Gavrilova, Marina
Depth-image-based rendering (DIBR) is a view synthesis technique that generates virtual views by warping from the reference images based on depth maps. The quality of synthesized views highly depends on the accuracy of depth maps. However, for dynamic scenarios, depth sequences obtained through stereo matching methods frame by frame can be temporally inconsistent, especially in static regions, which leads to uncomfortable flickering artifacts in synthesized videos. This problem can be eliminated by depth enhancement methods that perform temporal filtering to suppress depth inconsistency, yet those methods may also spread depth errors. Although these depth enhancement algorithms increase the temporal consistency of synthesized videos, they have the risk of reducing the quality of rendered videos. Since conventional methods may not achieve both properties, in this paper, we present for static regions a robust temporal depth enhancement (RTDE) method, which propagates exactly the reliable depth values into succeeding frames to upgrade not only the accuracy but also the temporal consistency of depth estimations. This technique benefits the quality of synthesized videos. In addition we propose a novel evaluation metric to quantitatively compare temporal consistency between our method and the state of arts. Experimental results demonstrate the robustness of our method for dynamic virtual view synthesis, not only the temporal consistency but also the quality of synthesized videos in static regions are improved.
A weight adjustment strategy to prevent cascade of boosted classifiers from overfitting
(Václav Skala - UNION Agency, 2015) Park, Ki-Yeong; Dong-Seok, Kim; Skala, Václav; Gavrilova, Marina
We propose a weight adjustment strategy to prevent a cascade of boosted classifiers from overfitting and to achieve an improved performance. In cascade learning, overfitting often occurs due to the iterative applications of bootstrapping. Since false positives that the previous classifier misclassifies are collected as negative examples through bootstrapping, negative examples more similar to positive examples are prepared as stages go on, and thus classifiers become tuned to the positive examples. When overfitting occurs, the classifier cascade shows performance degradation more in the detection rate than in the false alarm rate. In the proposed strategy, the imbalance between the detection rate and the false alarm rate is evaluated by computing the weight ratio of positive examples to negative examples and it is compensated by adjusting the weight ratio prior to boosting at each stage. Experimental results confirm the effectiveness of the proposed strategy. For experiments, face and pedestrian classifier cascades were trained by employing previous approaches and the proposed strategy. By employing the proposed strategy, the detection rate of classifier cascades was significantly improved for both face and pedestrian.
Content-aware re-targeting of discrete element layouts
(Václav Skala - UNION Agency, 2015) Hartmann, Stefan; Krüger, Björn; Klein, Reinhard; Skala, Václav; Gavrilova, Marina
Example-based modeling is an active line of research within the computer graphics community. With this work we propose a re-targeting scheme for polygonal domains containing a layout composed of discrete elements. Retargeting such domains is typically achieved employing a deformation to the initial domain. The goal of our approach is it to compute a novel layout within the deformed domain re-using the discrete elements from the initial layout. We show that the deformed interior of the initial domain can guide the layout algorithm that places the discrete elements. We evaluate our algorithm by re-targeting several challenging city blocks and the results confirm that generated layouts are visually similar to the original ones.
Projector-leap motion calibration for gestural interfaces
(Václav Skala - UNION Agency, 2015) Najman, Pavel; Zahrádka, Jiří; Zemčík, Pavel; Skala, Václav; Gavrilova, Marina
Calibration of a projector and a tracking device is an essential step for interaction with projected content using gestures. We propose a novel technique for calibration of a data projector and a Leap Motion sensor. Using the proposed approach, users can calibrate the system by touching few points on the screen plane and in the space above it. No printed patterns, reflective markers, or additional tools are needed. The calibration process involves two steps. In the first step, we collect finger positions which we then use in the second step to find the calibration matrix and projector position. We compared the accuracy and precision of the proposed method to the accuracy and precision of a capacitive touchscreen in a touch based interaction task. During the evaluation we measured the Euclidean distance between the displayed and touched points. The best average distance for our method was 1.23 mm which is comparable to 0.79 mm for touch screen. The experiments demonstrate that the technique is suitable for an interaction with user interface elements designed in the usual way.
Pose-specific pedestrian classificatiion using multiple features in par-infrared images
(Václav Skala - UNION Agency, 2015) Kim, Dong-Seok; Park, Ki-Yeong; Skala, Václav; Gavrilova, Marina
We present a multiple feature-based, pose-specific pedestrian classification approach to improve classification performance for fair-infrared (FIR) images. Using pose-specific classifiers and multiple features has proved to be beneficial in visible-spectrum-based classification systems; therefore, we adapt both to an FIR-based classification system. For pose-specific classifiers, we separate poses into sets of front/back and right/left poses and estimate the pose using template matching. For feature extraction, we use histograms of local intensity differences (HLID) and local binary patterns (LBP). Experiments showed that the proposed approaches improve the classification performance of a baseline HLID/linSVM approach.
Explorative analysis of 2D color maps
(Václav Skala - UNION Agency, 2015) Steiger, M.; Bernard, J.; Thum, S.; Mittelstädt, S.; Hutter, M.; Keim, D.; Kohlhammer, J.; Skala, Václav; Gavrilova, Marina
Color is one of the most important visual variables in information visualization. In many cases, two-dimensional information can be color-coded based on a 2D color map. A variety of color maps as well as a number of quality criteria for the use of color have been presented. The choice of the best color map depends on the analytical task users intend to perform and the design space in choosing an appropriate 2D color map is large. In this paper, we present the ColorMap-Explorer, a visual-interactive system that helps users in selecting the most appropriate 2D color map for their particular use case. ColorMap-Explorer also provides a library of many color map implementations that have been proposed in the scientific literature. To analyze their usefulness for different tasks, ColorMap-Explorer provides use case scenarios to allow users to obtain qualitative feedback. In addition, quantitative metrics are provided on a global (i.e. per color map) and local (i.e. per point) scale. ColorMap-Explorer enables users to explore the strengths and weaknesses of existing as well as user-provided color maps to find the best fit for their task. Any color map can be exported to be reused in other visualization tools. The code is published as open source software, so that the visualization community can use both the color map library and the ColorMap-Explorer tool. This also allows users to contribute new implementations.
Heterogeneous dataset acquisition for a continuously expandable benchmark (CEB)
(Václav Skala - UNION Agency, 2015) Krolla, Bernd; Stricker, Didier; Skala, Václav; Gavrilova, Marina
Ongoing research within the field of computer vision yielded a wide range of image based 3D reconstruction approaches. Starting years ago with low resolution RGB images as input, we face today a wide and fast growing range of available imaging devices to perform this task. To allow for a good comparability of resulting reconstructions, many different benchmarks and datasets have been made available. At the same time, we observe, that these benchmarks commonly address only a single capturing approach omitting the chance to compare against results of other acquisition methods. In contrast to such homogeneous benchmarks, we present in this work a heterogeneous benchmark, considering different acquisition devices to obtain our datasets. Besides these datasets, we furthermore provide reference data for download. To lastly keep track of the rapidly increasing number of different acquisition sensors, we opt to provide occasional updates of this benchmark within the future.
A simple and efficient feature descriptor for fast matching
(Václav Skala - UNION Agency, 2015) Hast, Anders; Sablina, Victoria A.; Kylberg, Gustaf; Sintorn, Ida-Maria; Skala, Václav; Gavrilova, Marina
A very simple but efficient feature descriptor is proposed for image matching/registration applications where invariance is not important. The descriptor length is only three times the height of the local region in which the descriptor is calculated, and experiments were conducted to compare it to the SURF descriptor. In addition, it is shown, how the sampling can be modified in order to obtain a rotation invariant descriptor, while still keeping it simple and efficient. Examples from stitching in microscopy and stereo processing of pairs of photographs are given to prove the concept.
Simultaneous absorption and environment light reconstruction in optical tomography problem
(Václav Skala - UNION Agency, 2015) V., Afanasyev; A., Ignatenko; A., Voloboy; Skala, Václav; Gavrilova, Marina
Classic tomography algorithms applied in optical tomography require the light source pre-calibration and do not allow refining the light map in tomography algorithm. This article shows an approach to environment light reconstruction during the ART algorithm execution. It makes the optical tomography scanning process more fast and simple, allowing to exclude the light calibration stage.
A GPU-accelerated augmented Lagrangian based L1-mean curvature Image denoising algorithm implementation
(Václav Skala - UNION Agency, 2015) Myllykoski, Mirko; Glowinski, Roland; Kärkkäinen, Tommi; Rossi, Tuomo; Skala, Václav; Gavrilova, Marina
This paper presents a graphics processing unit (GPU) implementation of a recently published augmented Lagrangian based L1-mean curvature image denoising algorithm. The algorithm uses a particular alternating direction method of multipliers to reduce the related saddle-point problem to an iterative sequence of four simpler minimization problems. Two of these subproblems do not contain the derivatives of the unknown variables and can therefore be solved point-wise without inter-process communication. In particular, this facilitates the efficient solution of the subproblem that deals with the non-convex term in the original objective function by modern GPUs. The two remaining subproblems are solved using the conjugate gradient method and a partial solution variant of the cyclic reduction method, both of which can be implemented relatively efficiently on GPUs. The numerical results indicate up to 33-fold speedups when compared against a single-threaded CPU implementation. The pointwise treated subproblem that takes care of the non-convex term in the original objective function was solved up to 76 times faster.
Evaluation of space partitioning data structures for nonlinear mapping
(Václav Skala - UNION Agency, 2015) E. V., Myasnikov; Skala, Václav; Gavrilova, Marina
Nonlinear mapping (Sammon mapping) is a nonlinear dimensionality reduction technique operating on the data structure preserving principle. Several possible space partitioning data structures (vp-trees, kd-trees and cluster trees) are applied in the paper to improve the efficiency of the nonlinear mapping algorithm. At the first step specified structures partition the input multidimensional space, at the second step space partitioning structure is used to build up the list of reference nodes used to approximate calculations. The further steps perform initialization and iterative refinement of the low-dimensional coordinates of objects in the output space using created lists of reference nodes. Analyzed space partitioning data structures are evaluated in terms of the data mapping error and runtime. The experiments are carried out on the well-known datasets.
An improved non-orthogonal texture warping for better shadow rendering
(Václav Skala - UNION Agency, 2015) Milet, Tomáš; Navrátil, Jan; Zemčík, Pavel; Skala, Václav; Gavrilova, Marina
In interactive applications, shadows are traditionally rendered using the shadow mapping algorithm. The disadvantage of the algorithm is limited resolution of depth texture which may lead to aliasing artifacts on shadow edges. This paper introduces an improved depth texture warping with non-orthogonal grid that can be employed for all kinds of light sources. For instance, already known approaches for reducing aliasing artifacts are widely used in outdoor scenes with directional light sources but they are not directly applicable for point light sources. We show that the improved warping parameterization reduces the aliasing artifacts and we are able to present high quality shadows regardless of a light source or a camera position in the scene.
Image-based object modeling by fitting salient lines and geometric primitives
(Václav Skala - UNION Agency, 2015) Meng-Hong, Cho; I-Chen, Lin; Skala, Václav; Gavrilova, Marina
With modern vision techniques and depth sensing devices, it becomes possible for common users to acquire the shape of an object from a set of color or depth images from different views. However, the estimated 3D volume or point clouds, disturbed by noise and errors, cannot directly be applied for graphics usage. This paper presents a two-stage method for reconstructing 3D graphics models from point clouds and photographs. Unlike related work that immediately fitted primitives for the point clouds, we propose finding the primary planes through salient lines in images in advance, and extracting auxiliary planes according to the symmetric properties. Then, a RANSAC method is used to fit primitives for the residual points. Intuitive editing tools are also provided for rapid model refinement. The experiments demonstrate that the proposed automatic stages can generate more accurate results. Besides, the user intervention time is less than that by a well known modeling tool.
Machine learning approach to automate facial expressions from physical activity
(Václav Skala - UNION Agency, 2015) Boukhalfi, Tarik; Desrosiers, Christian; Paquette, Eric; Skala, Václav; Gavrilova, Marina
We propose a novel approach based on machine learning to simulate facial expressions related to physical activity. Because of the various factors they involve, such as psychological and biomechanical, facial expressions are complex to model. While facial performance capture provides the best results, it is costly and difficult to use for real-time interaction during intense physical activity. A number of methods exist to automate facial animation related to speech or emotion, but there are no methods to automate facial expressions related to physical activity. This leads to unrealistic 3D characters, especially when performing intense physical activity. This research highlights the link between physical activity and facial expression, and to propose a data-driven approach providing realistic facial expressions, while leaving creative control. First, biological, mechanical, and facial expression data are captured. This information is then used to train regression trees and support vector machine (SVM) models, which predict facial expressions of virtual characters from their 3D motion. The proposed approach can be used with real-time, pre-recorded or key-framed animations, making it suitable for video games and movies as well.

Browse

Recent Submissions