WSCG 2016: Full Papers Proceedings

Permanent URI for this collection

http://hdl.handle.net/11025/26389

Browse

Showing 1 - 12 out of 12 results

CellPathway: a simulation tool for illustrative visualization of biochemical networks
(Václav Skala - UNION Agency, 2016) Reisacher, Matthias; Viola, Ivana; Le Muzic, Mathieu; Skala, Václav
The molecular knowledge about complex biochemical reaction networks in biotechnology is crucial and has received a lot of attention lately. As a consequence, multiple visualization programs have been already developed to illustrate the anatomy of a cell. However, since a real cell performs millions of reactions every second to sustain live, it is necessary to move from anatomical to physiological illustrations to communicate knowledge about the behavior of a cell more accurately. In this publication we propose a reaction system including a collision detection algorithm, which is able to work at the level of single atoms, to enable simulation of molecular interactions. To visually explain molecular activities during the simulation process, a real-time glow effect in combination with a clipping object have been implemented. Since intracellular processes are performed with a set of chemical transformations, a hierarchical structure is used to illustrate the impact of one reaction on the entire simulation. The CellPathway system integrates acceleration techniques to render large datasets containing millions of atoms in real-time, while the reaction system is processed directly on the GPU to enable simulation with more than 1000 molecules. Furthermore, a graphical user interface has been implemented to allow the user to control parameters during simulation interactively.
Cytological low-quality image segmentation using nonlinear regression, K-means and watershed
(Václav Skala - UNION Agency, 2016) Franco, Ramon A. S.; Martins, Paulo S.; Carvalho, Marco A.G. de; Skala, Václav
Since 1950, conventional cytology uses glass slides for microscopic analysis of cervical cells, in order to perform Pap Test. Such method yields low-quality images and overlapping cells, which both hampers their analysis and classification. Several countries use a modern method for the realization of Pap test called ThinPrep because it offers high- quality images and overcomes the problem of overlapping cells. ThinPrep facilitated the development of advanced image processing techniques for segmentation and classification of cervical cells. However, this method is not used by most of the developing countries of the world due to its relative high cost. This paper presents an algorithm for segmenting digital images obtained from conventional cytology method on glass slides. The technique usesWatershed Transform and K-Means Clustering in order to find cell markers or seeds. Nonlinear regression is applied as a way to refine the markers and to allow again the Watershed Transform utilization. We apply the technique in 10 glass slides of pap smears with a total of 67 cells. Our proposed technique has a promising performance in terms of accuracy of about 85%.
Optical flow estimation via steered-L1 norm
(Václav Skala - UNION Agency, 2016) Zayouna, Ammar; Comley, Richard; Shi, Daming; Skala, Václav
Global variational methods for estimating optical flow are among the best performing methods due to the subpixel accuracy and the ‘fill-in’ effect they provide. The fill-in effect allows optical flow displacements to be estimated even in low and untextured areas of the image. The estimation of such displacements are induced by the smoothness term. The L1 norm provides a robust regularisation term for the optical flow energy function with a very good performance for edge-preserving. However this norm suffers from several issues, among these is the isotropic nature of this norm which reduces the fill-in effect and eventually the accuracy of estimation in areas near motion boundaries. In this paper we propose an enhancement to the L1 norm that improves the fill-in effect for this smoothness term. In order to do this we analyse the structure tensor matrix and use its eigenvectors to steer the smoothness term into components that are ‘orthogonal to’ and ‘aligned with’ image structures. This is done in primal-dual formulation. Results show a reduced end-point error and improved accuracy compared to the conventional L1 norm.
Procedural texture synthesis by locally controlled spot noise
(Václav Skala - UNION Agency, 2016) Pavie, Nicolas; Gilet, Guillaume; Dischler, Jean-Michel; Ghazanfarpour, Djamchid; Skala, Václav
Procedural noises based on power spectrum definition and random phases have been widely used for procedural texturing, but using a noise process with random phases limits the types of possible patterns to Gaussian patterns (i.e. irregular textures with no structural features). Local Random Phase (LRP) Noise has introduced control over structural features in a noise model by fixing the frequencies and phase information of desired features, but this approach requires storing these frequencies. Space distortion and randomization must also be used to avoid repetitions and periodicity. In this paper, we present a noise model based on non-uniform random distributions of multiple Gaussian functions for synthesizing semi-structured textures. We extend the LRP noise model by using a spot noise based on a controlled distribution of kernels (spots), as an alternative formulation to local noises aligned on a regular grid. Spots are created as a combination of Gaussian functions to match either a specific power spectrum or a user-defined texture element. Our noise model improves the control over local structural features while keeping the benefits of LRP noise.
An EM based approach for motion segmentation of video sequence
(Václav Skala - UNION Agency, 2016) Zhao, Wei; Roos, Nico; Skala, Václav
Motions are important features for robot vision as we live in a dynamic world. Detecting moving objects is crucial for mobile robots and computer vision systems. This paper investigates an architecture for the segmentation of moving objects from image sequences. Objects are represented as groups of SIFT feature points. Instead of tracking the feature points over a sequence of frames, the movements of feature points between two successive frames are used. The segmentation of motions of each pair of frames is based on the expectation-maximization algorithm. The segmentation algorithm is iteratively applied over all frames of the sequence and the results are combined using Bayesian update.
Multiresolution Laplacian sparse coding technique for image representation
(Václav Skala - UNION Agency, 2016) Jemel, Intidar; Ejbali, Ridha; Zaied, Mourad; Skala, Václav
Sparse coding techniques have given good results in different domains especially in feature quantization and image representation. However, the major weakness of those techniques is their inability to represent the similarity between features. This limitation is due to the separate representation of features. Although the Laplacian sparse coding doesn’t focus on the spatial similarity in the image space, it preserves the locality of the features only in the data space. Due to this, the similarity between two local features belong to the similarity of their spatial neighborhood in the image. To overcome this flaw, we propose the integration of similarity based on Kullback-Leibler and wavelet decomposition in the domain of an image. This technique may surmount those limitations by taking into account each element of an image and its neighbors in similarity calculation. Classifications rates given by our approach show a clear improvement compared to those cited in this article.
Scene understanding using context-based conditional random field
(Václav Skala - UNION Agency, 2016) Zolghadr, Esfandiar; Furht, Borko; Skala, Václav
In this paper, a new framework for scene understanding using multi-modal high-ordered context-model is introduced. Spatial and semantical interactions are considered as sources of context which are incorporated in the model using a single object-scene relevance measure that quantifies high-order object relations. This score is used to minimize semantical inconsistencies among objects in dense graph representation of the scene category during the object recognition process. New context model is later incorporated in a Conditional Random Fields (CRF) framework to combine contextual cues with appearance descriptors in order to increase object localization and class prediction accuracy. A novel context-based non-central hypergeometric unary potential is defined to maximize the semantical coherence in the scene. Further refinement is performed using context-based pairwise and high-order potentials which use alpha-expansion and graph-cut to find optimal configuration. Comparison between the purposed approach and state-of-art algorithms shows effectiveness of this approach in annotation and interpretation of scenes.
Fracturing sparse-voxel-octree objects using dynamical Voronoi patterns
(Václav Skala - UNION Agency, 2016) Domaradzki, Jakub; Martyn, Tomasz; Skala, Václav
We introduce a new Voronoi-based method to fracture objects represented by sparse voxel octrees (SVOs). Our approach is inspired by the pattern-based methods, however, in contrast to them, it doesn’t require pattern precomputation. Moreover, thanks to the octree structure, the surfaces of the fractured pieces of geometry are created efficiently and robustly. Every fracture pattern is unique and centered at the impact location. A novel islands detection technique is also provided, which is tunable to a desired level-of-detail accuracy. The fractured pieces, which are determined as a consequence of the object’s destruction, are represented by individual SVOs, and treated and simulated as rigid bodies. For this purpose, we also propose a new collision detection technique, which extends the previous image-based methods to voxels. As a result, deep penetrations of colliding objects, resolved on various levels of physics that can be specified individually for each pair of the objects, are handled in parallel with no extra cost. In order to demonstrate our technique, a number of scenarios are presented, including a partial fracturing of objects with fine details.
Real-time voxel rendering algorithm based on screen space billboard voxel buffer with sparse lookup textures
(Václav Skala - UNION Agency, 2016) Jabłoński, Szymon; Martyn, Tomasz; Skala, Václav
In this paper, we present a novel approach to efficient real-time rendering of numerous high-resolution voxelized objects. We present a voxel rendering algorithm based on triangle rasterization pipeline with screen space rendering computational complexity. In order to limit the number of vertex shader invocations, voxel filtering algorithm with fixed size voxel data buffer was developed. Voxelized objects are represented by sparse voxel octree (SVO) structure. Using sparse texture available in modern graphics APIs, we create a 3D lookup table for voxel ids. Voxel filtering algorithm is based on 3D sparse texture ray marching approach. Screen Space Billboard Voxel Buffer is filled by voxels from visible voxels point cloud. Thanks to using 3D sparse textures, we are able to store high-resolution objects in VRAM memory. Moreover, sparse texture mipmaps can be used to control object level of detail (LOD). The geometry of a voxelized object is represented by a collection of points extracted from object SVO. Each point is defined by position, normal vector and texture coordinates. We also show how to take advantage of programmable geometry shaders in order to store voxel objects with extremely low memory requirements and to perform real-time visualization. Moreover, geometry shaders are used to generate billboard quads from the point cloud and to perform fast face culling. As a result, we obtained comparable or even better performance results in comparison to SVO ray tracing approach. The number of rendered voxels is limited to defined Screen Space Billboard Voxel Buffer resolution. Last but not least, thanks to graphics card adapter support, developed algorithm can be easily integrated with any graphics engine using triangle rasterization pipeline.
Accelerating spatial data structures in ray tracing through precomputed line space visibility
(Václav Skala - UNION Agency, 2016) Keul, Kevin; Müller, Stefan; Lemke, Paul; Skala, Václav
We propose an efficient approach to precompute and reuse visibility information based on existing spatial data structures by using a precomputed data structure: the line space. This data structure provides an additional skip condition by checking whether the subnodes in a hierarchical spatial data structures need to check for intersection with the ray. We evaluate this method on different test scenes and show that it is able to achieve a remarkable speed-up by using this skip condition. Furthermore we describe algorithms for fast set-up and traversal in detail and discuss important strategies for this approach.
Identifying linear vector fields on 2D manifolds
(Václav Skala - UNION Agency, 2016) Volke, Sebastian; Koch, Stefan; Hlawitschka, Mario; Skala, Václav
Local linearity of vector fields is a property that is well researched and understood. Linear approximation can be used to simplify algorithms or for data reduction. Whereas the concept is easy to implement in 2D and 3D, it loses meaning on manifolds as linearity has either to be defined based on an embedding in a higher-dimensional Cartesian space or on a map. We present an adaptive atlas-based vector field decomposition to solve the problem on manifolds and present its application on synthetic and climate data.
Using trajectories derived by dense optical flows as a spatial component in background subtraction
(Václav Skala - UNION Agency, 2016) Radolko, Martin; Farhadifard, Fahimeh; Skala, Václav
Foreground-Background Segregation has been intensively researched in the last decades as it is an important first step in many Computer Vision tasks. Nonetheless, there are still many open questions in this area and in this paper we focus on a special surveillance scenario where a static camera monitors a predefined region. This restrain makes some aspects easier and good results could be achieved with Background Subtraction methods. However, these only work pixelwise and lack the spatial component completely. We suggest an approach to add the crucial spatial information to the segmentations with Dense Optical Flows. For this, a number of successive images are taken from the video to compute the Trajectories of the pixels through these frames. This enables us to fuse the information from the several images and use this for segmentation. The algorithm was evaluated on a video from a surveillance camera and showed promising results.

Browse

Recent Submissions