WSCG 2017: Poster Papers Proceedings
Permanent URI for this collection
Browse
Recent Submissions
Item Semantic search user interface patterns: an introduction(Václav Skala - UNION Agency, 2017) Marx, Edgar; Khalili, Ali; Valdestilhas, André; Skala, VáclavWithin the past few years, many patterns and principles have been proposed towards the enhancement of search user interfaces and experience. However, to access and explore information efficiently is still significantly challenging. Recently, we have seen the rise of a new kind of information retrieval approach, the so-called semantic search systems. These systems promise more accurate results while exploring semantics of the data. Although there exist several search user interfaces tailored to semantic search, there is still a lack of usability studies as well as good practices. In this work, we discuss the applicability of traditional search user interfaces in semantic search systems. Furthermore, we propose a new interaction model based on four patterns: Poli-Communicative, Discrete Display, Heterogeneous Data-face and Dive in-place.
Item Multi-references shape constraint for snakes(Václav Skala - UNION Agency, 2017) Amine, Mohammed; Sakly, Ines; Mhiri, Slim; Ghorbel, Faouzi; Skala, VáclavIn this research, we intend to present a new method of snakes with an invariant shape prior. We consider the general case where different templates are available and we have to choose the most suitable ones to define the shape constraint. A new external force is then proposed which is able to take into account several references at the same time with proportional weighting factors. Both a Fourier based shape alignment method and a complete and stable set of shape descriptors are used to ensure invariance and robustness of the prior knowledge to Euclidean transformations. To illustrate the efficiency of our approach, a set of experiments are applied on synthetic and real data. Promising results are obtained and commented.Item Scene text segmentation based on redundant representation of character candidates(Václav Skala - UNION Agency, 2017) Saric, Matko; Skala, VáclavText segmentation is important step in extraction of textual information from natural scene images. This paper proposes novel method for generation of character candidate regions based on redundant representation of subpaths in extremal regions (ER) tree. These subpaths are constructed using area variation and pruned using their length: each sufficiently long subpath is character candidate which is represented by subset of regions contained in the subpath. Mean SVM probability score of regions in subset is used to filter out non character components. Proposed approach for character candidates generation is followed by character grouping and restoration steps. Experimental results obtained on the ICDAR 2013 dataset shows that the proposed text segmentation method obtains second highest precision and competitive recall rate.Item An interactive Tunisian virtual museum through affine reconstruction of gigantic mosaics and antic 3-D models(Václav Skala - UNION Agency, 2017) Elghoul, S.; Saidani, M.; Ghorbel, F.; Skala, VáclavMuseums are no longer static depositories for objects, as they used to be for the past two centuries. Online and interactive access has created new opportunities for museums and cultural institutions to reach out and discover new audiences for promotion of cultural achievements. The virtual tours and the panorama reconstructions become among the most popular solutions for virtual consultation of historical monuments. A Tunisian virtual museum application aims to make its relative monuments more accessible to experts such as archaeologist or art historians, and even to the large public. In this research, we intend to present an invariant based approaches for the reconstruction of both mosaic panorama and 3D models. We propose a framework to create a Tunisian virtual museum and we focused on an interactive application related to Bardo museum. We apply affine invariance in a finite set of viewpoints to compute shape descriptors with completeness and stability properties. Such affine invariants serve to refine 3D model generated by classic shape-from-silhouette algorithm and to assist the creation of panorama image mosaics from uncalibrated images.Item Isomorphic loss function for head pose estimation(Václav Skala - UNION Agency, 2017) Felea, Iulian; Florea, Corneliu; Vertan, Constantin; Florea, Laura; Skala, VáclavAccurate head pose estimation is a key step in many practical applications involving face analysis tasks, such as emotion recognition. We address the problem of head pose estimation in still color images acquired with a standard camera with limited resolution details. To achieve the proposed goal, we make use of the recent advances of Deep Convolutional Neural Networks. As head angles with respect on yaw and pitch are continuous, the problem is one of regression. Typical loss function for regression are based on L1 and L2 distances which are notorious for susceptibility to outliers. To address this aspect we introduce an isomorphic transformation which maps the initially infinite space into a closed space compressed at the ends and thus significantly down–weighting the significance of outliers. We have thoroughly evaluated the proposed approach on multiple publicly head pose databases.Item Convolutional neural network based chart image classification(Václav Skala - UNION Agency, 2017) Amara, Jihen; Kaur, Pawandeep; Owonibi, Michael; Bouaziz, Bassem; Skala, VáclavCharts are frequently embedded objects in digital documents and are used to convey a clear analysis of research results or commercial data trends. These charts are created through different means and may be represented by a variety of patterns such as column charts, line charts and pie charts. Chart recognition is as important as text recognition to automatically comprehend the knowledge within digital document. Chart recognition consists on identifying the chart type and decoding its visual contents into computer understandable values. Previous work in chart image identification has relied on hand crafted features which often fails when dealing with a large amount of data that could contain significant varieties and less common char types. Hence, as a first step towards this goal, in this paper we propose to use a deep learning-based approach that automates the feature extraction step. We present an improved version of the LeNet [LeCu 89] convolutional neural network architecture for chart image classification. We derive 11 classes of visualization (Scatter Plot, Column Chart, etc.) which we use to annotate 3377 chart images. Results show the efficiency of our proposed method with 89.5 % of accuracy rate.Item Volumetric ultrasound imaging: modeling of waveform inversion(Václav Skala - UNION Agency, 2017) Belykh, Igor; Markov, Dmitry; Skala, VáclavUltrasound imaging is widely used in medicine and nondestructive testing. Medical ultrasound volumetric imaging can be considered as a competitive method to X-ray CT and MRI. Recent results in breast ultrasound tomography provide spatial distribution of sound speed and attenuation coefficient based on inverse problem solution in frequency domain. Breast tissue can be characterized as relatively homogeneous medium with possible number of small diagnostically important target inhomogeneities (lesions). This paper is focused on numerical model of complex structure with objects of different sizes and properties that can be met in medicine or in various industrial or scientific cases. High resolution ultrasound imaging is sensitive to a number of effects such as reverberation, diffraction and attenuation that can degrade image quality and cause the artifacts. Those effects are investigated and eliminated by iterative waveform inversion performed in time domain for volumetric visualization of objects’ structure and acoustic properties.Item Efficient modeling methods of large-scale model for Monte Carlo transport simulation(Václav Skala - UNION Agency, 2017) Qin, Guiming; Ma, Yan; Fu, Yuanguang; Skala, VáclavMonte Carlo methods are widely used in simulation of the full-core reactor. They are usually adopted to deal with the geometry model based on the Constructive Solid Geometry. A visual modeling software for the automatic particle transport program, called JLAMT, is developed by the institute of applied physics and computation mathematics. It provides the computing model for Monte Carlo simulation codes, such as Geant4(a software toolkits developed by CERN) and JMCT(a 3D Monte Carlo neutron and photon transport code). To get a better result, the detailed model is needed. For devices as complex as full-core reactors, tens of thousands solids are needed to represent the model. This paper brings up efficient modeling methods of implicit modeling and layer-based modeling for solving this problem. And the effects to the overlap checking are discussed. Taking the full-core reactor of Daya bay power station as an example, experiments show that, by using the efficient modeling methods, both the amount of solids and the time of the overlap checking are reduced.Item An automatic hole filling method of point cloud for 3D scanning(Václav Skala - UNION Agency, 2017) Muraki, Yuta; Nishio, Koji; Kanaya, Takayuki; Kobori, Ken-ichi; Skala, VáclavIn recent years, due to the development of three-dimensional scanning technology, the opportunities for real objects to be three-dimensionally measured, taken into the PC as point cloud data, and used for various contents are increasing. However, the point cloud data obtained by three-dimensional scanning has many problems such as data loss due to occlusion or the material of the object to be measured, and occurrence of noise. Therefore, it is necessary to edit the point cloud data obtained by scanning. Particularly, since the point cloud data obtained by scanning contains many data missing, it takes much time to fill holes. Therefore, we propose a method to automatically filling hole obtained by three-dimensional scanning. In our method, a surface is generated from a point in the vicinity of a hole, and a hole region is filled by generating a point sequence on the surface. This method is suitable for processing to fill a large number of holes because point sequence interpolation can be performed automatically for hole regions without requiring user input.Item Efficient atmospheric rendering for mobile virtual globes(Václav Skala - UNION Agency, 2017) Santana, José M.; Trujillo, Agustín; Suárez, José P.; Ortega, Sebastián; Skala, VáclavMost virtual globe systems feature a rendering of the atmosphere that surrounds the earth. This element is so widespread that seeing a virtual earth without a surrounding halo makes the image seem much more artificial. Atmospheric rendering is a costly process aimed to enhance the realism and beauty of the scene. For that reason many virtual globe systems, specially in low-resourced devices, rely on simplified schemes to represent the atmosphere. However, the accurate representation of the atmosphere implies the volumetric rendering of the semi-transparent air mass that covers the whole geographical scene. Thus, the color of each pixel is composed of the light scattered by all the points in space projected on that pixel. The present work takes advantage of the spherical symmetry of the atmosphere’s mathematical model to implement efficiently this volumetric rendering on mobile devices.Item Background modeling: dealing with pan, tilt or zoom in videos(Václav Skala - UNION Agency, 2017) Radolko, Martin; Farhadifard, Fahimeh; Freiherr von Lukas, Uwe; Skala, VáclavEven simple camera movements like pan, tilt or zoom constitute enormous problems for background subtraction algorithms since the modeling of the background works only under the assumption of a static camera. The problem has been mostly ignored and other algorithms have been used for videos with non-static cameras. Nonetheless, in this paper we introduce a method that adapts the background model to these camera movements by using affine transformations in combination with a similarity metric, and thereby the algorithm makes background subtraction usable for these situations. Also, to keep the generality of this approach, we first apply a detection step to avoid unnecessary adaptions in videos with a static camera because even small adaptions might otherwise deteriorate the background model over time. The method is evaluated on the extensive changedetection.net data set and could reliably detect camera motion in all videos as well as precisely adapt the model of the background to that motion. This does improve the quality of the background models significantly which consequently leads to a higher accuracy of the segmentations.Item Fashion recommendations using text mining and multiple content attributes(Václav Skala - UNION Agency, 2017) Zhou, Wei; Zhou, Yanghong; Li, Runze; Mok, P. Y.; Skala, VáclavMany online stores actively recommend commodities to users for facilitating easy product selection and increasing product exposure. Typical approach is by collaborative filtering, namely recommending the products based on their popularity, assuming that users may buy the products that many others have purchased. However, fashion recommendation is different from other product recommendations, because people may not like to go with the crowd in selecting fashion items. Other approaches of fashion recommendations include providing suggestions based on users’ purchase or browsing history. This is mainly done by searching similar products using commodities’ tags. Yet, the accuracy of tag-based recommendations may be limited due to ambiguous text expression and nonstandard tag names for fashion items. In this paper we collect a large fashion clothing dataset from different online stores. We develop a fashion keyword library by statistical natural language processing, and then we formulate an algorithm to automatically label fashion product attributes according to the defined library by text mining and semantic analysis. Lastly, we develop novel fashion recommendation models to select similar and mix-and-match products by integrating text-based product attributes and image extracted features. We evaluate the effectiveness of our approach by experiment over real datasets.Item Improved adaptive background subtraction method using pixel-based segmenter(Václav Skala - UNION Agency, 2017) Batuhan Baskurt, Kemal; Same, Refik; Skala, VáclavMoving object detection is essential in many computer vision systems as it is generally first process which feeds following algorithmic steps after getting camera stream. Thus quality of moving object detection is crucial for success of the whole process flow. It has been studied in the literature over the last two decades but it is still challenging issue because of factors such as background complexity, illumination variations, noise, occlusion and run-time performance requirement considering rapidly increasing image size and quality. In this paper, we try to contribute to solve this problem by improving an existing realtime non-parametric moving object detection method. In scope of this study, pixel based background model in which each pixel is represented separately by its distribution on time domain is used. Mentioned discrete background model is suitable for parallel processing by dividing the image to sub images in order to accelerate the process. Main feature of proposed nonparametric approach is automatic adjustment of algorithm parameters according to changes on the scene. This feature provides easy adaptation to environmental change and robustness for different scenes with unique parameter initialization. Another contribution is scene change detector to handle sudden illumination changes and adopt the background model to new scene in the fastest way. Experiments on 2012 ChangeDetection.net dataset show that our approach outperforms most state-of-the-art methods. Improvement obtained both on robustness and practical performance provides our approach to be able to use in real world monitoring systems.Item Ellipse detection in very noisy environment(Václav Skala - UNION Agency, 2017) Zoppitelli, Pierre; Mavromatis, Sébastien; Sequeira, Jean; Skala, VáclavEllipse detection is a major issue of image analysis because circles are transformed into ellipses by projective transformations, and most of 3D scenes contain circles that are significant for understanding them (mechanical parts, man-made objects, interior decoration,). Several algorithms have been designed and published, some of them very recently, to characterize ellipses within images and they are very efficient in most classical situations. But they all fail in some specific cases that regularly happen as, for example, when the noise in the image is such as it produces dashed ellipses, or when they are very flat, or when only small parts of them are visible. We propose an algorithm that brings a solution in such cases, even if it is not more efficient than the other ones in classical situations. Hence, it can be used as a complement of other algorithms when we want to detect ellipses in a robust way, i.e. in all situations. This algorithm takes advantage of a property of ellipses related to their tangent lines, without any assumption on edge connectivity: primitives are designed to characterize the possibility for a point and an orientation to locally represent an edge; these primitives are not connected and their global analysis enables to obtain the center location and the three other parameters of ellipses that can be drawn through this set of primitives, i.e. that go through some of these points and that have the corresponding tangent lines. A set of tests has been used to measure its robustness.Item Evaluation of an object detection system in the submarine environment(Václav Skala - UNION Agency, 2017) Rekik, Farah; Ayedi, Walid; Jallouli, Mohammed; Skala, VáclavThe object detection in underwater environment requires a perfect description of the image with appropriate features, in order to extract the right object of interest. In this paper we adopt a novel underwater object detection algorithm based on multi-scale covariance descriptor (MSCOV) for the image description and feature extraction, and support vector machine classifier (SVM) for the data classification. This approach is evaluated in pipe detection application using MARIS dataset. The result of this algorithm outperforms existing detection system using the same dataset. Computer vision in underwater environment suffers from absorption and scattering of light in water. Despite the work carried out so far, image preprocessing is the only solution to cope with this problem. This step creates a waste of time and requires hardware and software resources. But the proposed method does not require pretreatment so it accelerate the process.Item A new dimensionality reduction-based visualization approach for massive data(Václav Skala - UNION Agency, 2017) Paulauskienė, Kotryna; Kurasova, Olga; Skala, VáclavWe live in a big data and data analytics era. The volume, velocity, and variety of data generated today require special methods and techniques for data analysis and inferencing. Data visualization tools allow us to understand the data deeper. One of the straightforward ways of multidimensional data visualization is based on dimensionality reduction and illustrated by a scatter plot. However, visualization of millions of points in a scatter plot does not make a sense. Usually, data sampling or clustering is performed before visualization to reduce the amount of the visualized points, but in such a case, meaningful outliers can be rejected and will not be visualized. In this paper, a new approach for massive data visualization without point overlapping is proposed and investigated. The approach consists of two main stages: selection of a data subset and its visualization without overlapping. The experiments have been carried out with ten data sets. The efficiency of subset selection and visualization of data subset projection is confirmed by a comprehensive set of comparisons.Item Video summarization based on local features(Václav Skala - UNION Agency, 2017) Massaoudi, Mohammed; Bahroun, Sahbi; Zagrouba, Ezzeddine; Skala, VáclavKeyframe extraction process consists on presenting an abstract of the entire video with the most representative frames. It is one of the basic procedures relating to video retrieval and summary. This paper present a novel method for keyframe extraction based on SURF local features. First, we select a group of candidate frames from a video shot using a leap extraction technique. Then, SURF is used to detect and describe local features on the candidate frames. After that, we analyzed those features to eliminate near duplicate keyframes, helping to keep a compact set, using FLANN method. We developed a comparative study to evaluate our method with three state of the art approaches based on local features. The results show that our method overcomes those approaches.Item Evaluation of standard and directional median filters algorithms and their implementation on FPGA for medical application(Václav Skala - UNION Agency, 2017) Talbi, F.; Alim-Ferhat, F.; Seddiki, S.; Hachemi, B.; Skala, VáclavThe standard and the directional median filters are effective methods for the removal of impulse-based noise from the images. The main advantage is being the preserving of edges as compared to the mean filter. The main objective of this article is to implement the standard and the directional median filters on FPGA (Field Programmable Gate Array) in order to eliminate impulsive noise in the medical image. As a first step, two algorithms were developed and validated by Matlab tool. Subsequently, two architectures are proposed and implemented using the Xilinx ISE 12.2 environment.Item Ray tracing API integration for OpenGL applications(Václav Skala - UNION Agency, 2017) Lai, Wei-Hao; Tang, Chang-Yu; Chang, Chun-Fa; Skala, VáclavRay tracing is one of the most important rendering techniques in computer graphics. By means of simulating reflection and refraction of light transportation, ray tracing generates more photorealistic images than scanline rendering. However, the high computational cost is the main disadvantage of ray tracing algorithm. In recent years, the computing power of GPU has increased dramatically, and general-purpose computing on graphics processing units (GPGPU) has become popular. Many scholars have presented some physically based rendering methods with CUDA or OpenCL in order to improve image quality and increase rendering speed. Because rasterization is the mainstream in the gaming industry, there is still a long way to go to make ray tracing accepted by the industry in the near future. We introduce a ray tracing API integration for OpenGL applications that can replace the original OpenGL rasterization with ray tracing by simply adding a few lines of code, and the ray tracing algorithm in this API is parallelized by OpenCL.
