WSCG 2018: Poster Papers Proceedings
Permanent URI for this collection
Browse
Recent Submissions
Item Сompression for texture images in different basis functions using system criteria analysis(Václav Skala - UNION Agency, 2018) Babikov, A.Yu.; Voronin, V. V.; Ryzhov, V. P.; Ryzhov, Yu.V.; Skala, VáclavThis paper considers image compression for texture images. For texture representation we consider the orthogonal decomposition of two-dimensional signals (images) using spectral transform in the different basis functions. This paper focuses on the analysis of the following basis DCT, FFT, Haar, Hartley, and Walsh using system criteria analysis. The error of the orthogonal representation of images and the computational cost are considered when choosing a basis system, which is based on the Bellman-Zadeh concept using fuzzy sets. It is shown that the Haar transform can represent textural images more efficiently with smaller average risk than other basis functions.
Item The solution of the problem of simplifying the images for the subsequent minimization of the image bit depth(Václav Skala - UNION Agency, 2018) Semenishchev, Evgenii; Voronin, Viacheslav; Shraifel, Igor; Skala, VáclavIn this paper, the approach of changing bit depth of images is considered. This type of operation is required when performing primary processing operations, identifying parameters and stitching images. The process of changing bits depth of images is performed in three stages. At each stage, the error minimization criterion is tested Result of applying the approach allows obtaining numerical region characteristics including the number of clusters, the number of minimum and maximum cluster sizes. To perform the process of minimizing some of the criteria, it is necessary to divide the image into areas. The paper presents a mathematical description of the approach, as well as flowcharts for performing operations of data processing steps. The article gives recommendations for choosing coefficients to obtain optimal minimizing parameters. The test images give an example of performing bit changes on image areas.Item Using a spatiotemporal plane to recover a moving object’s shape using Spatiotemporal Boundary Formation(Václav Skala - UNION Agency, 2018) Cunningham, Douglas W.; Skala, VáclavWhen ever an object moves, it successively covers and uncovers surfaces that are farther away. This occlusion and dis-occlusion always occurs precisely at the boundaries of the moving object and as such provide information not only about the shape of the object but also about its velocity, transparency, and relative depth. Humans can and do use this information, and the process has come to be called Spatiotemporal Boundary Formation (SBF). Previous authors have used the wealth of experimental investigations into SBF to create a mathematical model of the process. In this article we proposed a novel method to recover the orientation and velocity the local edge segments of the moving objects which is more flexible, more robust, more compact, and allows the recovery of edges that do not have a constant velocity. The method can be used in object segmentation algorithms or as a pre-filter for machine-learning-based recognition algorithms in order to improve the overall result.Item The fast semi-bounded kernel-diffeomorphism estimator(Václav Skala - UNION Agency, 2018) Ben Othman, Ibtissem; Troudi, Molka; Ghorbel, Faouzi; Skala, VáclavWe introduce, by this work, a fast method to estimate probability density functions in the semi-bounded case. This new technique is a simplified version of the kernel-diffeomorphism estimator which requires complexity in the calculations. It is based on a logarithmic transformation of the data which will be estimated by the conventional kernel estimator. Thus, the algorithm complexity is reduced from O(N2) to O(N).Item A persistent naming system based on graph transformation rules(Václav Skala - UNION Agency, 2018) Marcheix, David; Cardot, Anaïs; Skapin, Xavier; Dieudonné-Glad, Nadine; Skala, Václav3D modeling for Archaeology requires to easily model scenes by letting users evaluate a parametric specification of archaeology-oriented gestures, then modify and reevaluate the specification to produce various restitution hypotheses. But the current modeling tools that support reevaluation mechanisms are not dedicated to Archaeology. The Jerboa library, based on graph transformations rules, is well suited for creating operations fitting the needs of archaeologists. But it does not any support reevaluation mechanism and especially the persistent naming system, that is used to identify the entities of the initial model and match them with entities of the reevaluated model. In this paper, we extend Jerboa with a new application-independent persistent naming model, which is more general and homogeneous than other solutions found in the literature and is the first one to handle parametric specification edition.Item Deep learning for historical cadastral maps digitization: overview, challenges and potential(Václav Skala - UNION Agency, 2018) Ignjatić, Jelena; Nikolić, Bojana; Rikalović, Aleksandar; Skala, VáclavCartographic heritage of historical cadastral maps represent remarkable geospatial data. Historical cadastral maps are generally regarded as an essential part of the land management infrastructure (buildings, streets, canals, bridges, etc.). Today these cadastral maps are still in use in a digital raster form (scanned maps). Digitization of cadastral maps is time consuming and it is a challenge for scientists and engineers to find ways to automatically convert raster into vector maps. The process of map digitization typically involves several stages: preprocessing, visual object detection and classification, vector representation postprocessing and extracting information from text. Although neural networks have had a long history of use in the domain, their applications remain limited to extracting the information from text. Recent convergence of advancements in the domains of training deep neural networks (DNN) and GPU hardware allowed DNNs to achieve state-of-the-art results in computer vision applications, beyond hand-written text recognition. This paper provides an overview of different approaches to historical cadastral maps digitization, focusing of the challenges and the potential of using deep neural networks in map digitization.Item Time series social network visualization based on dimension reduction(Václav Skala - UNION Agency, 2018) Al-Ghalibi, Maha; Al-Azzawi, Adil; Skala, VáclavSocial networks are in general dynamically due to the involvement of many people on the web such as Facebook, Twitter, and Snapchat, etc. The meaningful visualization and analysis of social network is challenging due to its dynamic nature, the mobility of nodes in the network and extremely large size. In this paper, we consider the higher dimensionality issue of social networks regarding time series social network construction and visualization. To solve this issue, we develop a statically data-mining based approach for dimensionality reduction in social networks. Basically, we find that each sub-social network’s model has different dimensions by nodes and links which are sampled originally from an m-dimensional metric space. Experimentally, we find that the m-dimensional features for each sub-network cause fail connections in time-series during the network reconstruction model for visualization. Therefore, we propose a new dimension reduction approach that is based on developing an SVD algorithm by relying on select significant sub features. Then we extract time features from the feature space of the original dataset to visualize the network in a deferent time interval. However, to monitor the network development and also the dimensionality reduction of features help us to speed up the computation time of the shortest path. The social circle Facebook dataset form Stanford is used with its corresponding attributes. The dataset includes node features (profile), circles, and ego networks. The obtained result shows better performances regarding the computation time and network visualization. Moreover, the experimental results show that the proposed system is much faster than the approach based on the whole feature space for closeness centrality computing.Item Realistic lens distortion rendering(Václav Skala - UNION Agency, 2018) Lambers, Martin; Sommerhoff, Hendrik; Kolb, Andreas; Skala, VáclavRendering images with lens distortion that matches real cameras requires a camera model that allows calibration of relevant parameters based on real imagery. This requirement is not fulfilled for camera models typically used in the field of Computer Graphics. In this paper, we present two approaches to integrate realistic lens distortions effects into any graphics pipeline. Both approaches are based on the most widely used camera model in Computer Vision, and thus can reproduce the behavior of real calibrated cameras. The advantages and drawbacks of the two approaches are compared, and both are verified by recovering rendering parameters through a calibration performed on rendered images.Item Human action recognition based on 3D convolution neural networks from RGBD videos(Václav Skala - UNION Agency, 2018) Al-Akam, Rawya; Paulus, Dietrich; Gharabaghi, Darius; Skala, VáclavHuman action recognition with color and depth sensors has received increasing attention in image processing and computer vision. This paper target is to develop a novel deep model for recognizing human action from the fusion of RGB-D videos based on a Convolutional Neural Network. This work is proposed a novel 3D Convolutional Neural Network architecture that implicitly captures motion information between adjacent frames, which are represented in two main steps: As a First, the optical flow is used to extract motion information from spatio-temporal domains of the different RGB-D video actions. This information is used to compute the features vector values from deep 3D CNN model. Secondly, train and evaluate a 3D CNN from three channels of the input video sequences (i.e. RGB, depth and combining information from both channels (RGB-D)) to obtain a feature representation for a 3D CNN model. For evaluating the accuracy results, a Convolutional Neural Network based on different data channels are trained and additionally the possibilities of feature extraction from 3D Convolutional Neural Network and the features are examined by support vector machine to improve and recognize human actions. From this methods, we demonstrate that the test results from RGB-D channels better than the results from each channel trained separately by baseline Convolutional Neural Network and outperform the state of the art on the same public datasets.Item A new model driven architecture for deep learning-based multimodal lifelog retrieval(Václav Skala - UNION Agency, 2018) Ben Abdallah, Fatma; Feki, Ghada; Ben Ammar, Anis; Ben Amar, Chokri; Skala, VáclavNowadays, taking photos and recording our life are daily task for the majority of people. The recorded information helped to build several applications like the self-monitoring of activities, memory assistance and long-term assisted living. This trend, called lifelogging, interests a lot of research communities such as computer vision, machine learning, human-computer interaction, pervasive computing and multimedia. Great effort have been made in the acquisition and the storage of captured data but there are still challenges in managing, analyzing, indexing, retrieving, summarizing and visualizing these captured data. In this work, we present a new model driven architecture for deep learning-based multimodal lifelog retrieval, summarization and visualization. Our proposed approach is based on different models integrated in an architecture established on four phases. Based on Convolutional Neural Network, the first phase consists of data preprocessing for discarding noisy images. In a second step, we extract several features to enhance the data description. Then, we generate a semantic segmentation to limit the search area in order to better control the runtime and the complexity. The second phase consist in analyzing the query. The third phase which based on Relational Network aims at retrieving the data matching the query. The final phase treat the diversity-based summarization with k-means which offers, to lifelogger, a key-frame concept and context selection-based visualization.Item Human action recognition from RGBD videos based on retina model and local binary pattern features(Václav Skala - UNION Agency, 2018) Al-Akam, Rawya; Al-Darraji, Salah; Paulus, Dietrich; Skala, VáclavHuman action recognition from the videos is one of the most attractive topics in computer vision during the last decades due to wide applications development. This research has mainly focused on learning and recognizing actions from RGB and Depth videos (RGBD). RGBD is a powerful source of data providing the aligned depth information which has great ability to improve the performance of different problems in image understanding and video processing. In this work, a novel system for human action recognition is proposed to extract distinctive spatio and temporal feature vectors for presenting the spatio-temporal evolutions from a set of training and testing video sequences of different actions. The feature vectors are computed in two steps: The First step is the motion detection from all video frames by using spatio-temporal retina model. This model gives a good structuring of video data by removing the noise and illumination variation and is used to detect potentially salient areas, these areas represent the motion information of the moving object in each frame of video sequences. In the Second step, because of human motion can be seen as a type of texture pattern, the local binary pattern descriptor (LBP) is used to extract features from the spatio-temporal salient areas and formulated them as a histogram to make the bag of feature vectors. To evaluate the performance of the proposed method, the k-means clustering, and Random Forest classification is applied on the bag of feature vectors. This approach is demonstrated that our system achieves superior performance in comparison with the state-of-the-art and all experimental results are depending on two public RGBD datasets.
