Srubar,S.
Abstract:
Two different segmentations of the same image can be evaluated in many ways. One resulting number hardly generalizes all differences in segmentations. Moreover, common methods can evaluate similar segmentations as quite different. Proposed evaluation divides dissimilarity into granularity and border difference. Granularity difference represents number of segments while border difference evaluates a rate of agreement in delineating of objects in the image. Such approach evaluates segmentations more precisely and has natural meaning of both resulting values.
Zuzanak,J., Zemcik,P.
Abstract:
Graph rewriting systems are applicable to vast majority of problems that are being solved in computer science. From problems concerning program optimization, software verification, description, and parsing of structured information to graph programming languages and layout algorithms. Graph rewriting systems are often represented as sets of rules describing transformations on graphs. The graph rewriting rule encapsulates complete information about applicable graph modification. In context of the described graph rewriting system, rules represent atomic modification of graph. A novel approach to graph rewriting and criteria for rule application enabling development of exhaustive graph rewrite system is introduced. The presented approach is derived from the well known double pushout approach (DPO). This paper concentrates on discussion of knowledge formalization representation for modeling concepts and on application of these concepts using the proposed programmed graph rewriting system.
Sukilovic,T., Vukmirovic,S.
Abstract:
In this paper notion of discrete curvature associated to a black pixel of black and white raster image is defined. Possible areas of its application include detection of arbitrary shapes, especially those consisting of piecewise smooth curves. Using the discrete curvature approach, we describe and implement an algorithm for circle and circular arc detection, and compare it to the generalized Hough transform.
Diaz-Tula,A., Castaneda-Garay,M., Belmonte-Fernandez,O.
Abstract:
Under certain circumstances it is necessary to read the visible frame buffer from a display device for post-processing. Examples of such applications are systems where the user interacts through projection screens and cameras. Traditionally the frame buffer is copied to the RAM to perform the post-processing, which many times is parallelisable and compute-intensive, leading to high latency times associated with the transfer of the pixels information over the bus and serial execution on CPU. But to the purpose of processing the frame buffer copy with the programmable hardware of the graphics processing unit (GPU), it needs to be copied to Video RAM instead of conventional RAM to avoid the latency time and exploit the parallel computational resources of the GPU. On this work we present an easy-to-use approach for reading the visible frame buffer to a Pixel Buffer Object, suitable for processing with the GPU through the CUDA architecture.
Zabiniako,V., Rusakov,P.
Abstract:
This article provides recommendations for implementation of lightweight graph visualization system. Comparison of according existing freeware and shareware solutions is being made. Overall proposed software architecture at high abstraction level is presented along with details of implementation of each mechanism. This work includes aspects of usage of XML format for improved graph data storing and parsing activities; optimization of force-based graph layout algorithm (based on original work by Peter Eades); description of useful visualization techniques (such as transparency, illumination distance, projective shadows, etc.). Additional capability that might be useful in database design and analysis routines: visual data clustering is introduced with theoretical description of according methodologies. Described ideas were implemented and verified in original lightweight software 3DIIVE. Conclusions about achieved results and further work are also presented.
Chang,Y.F., Chen, S.Y.
Abstract:
Based on cloud computing and software service, the “computer-aided design interface for smart home device” helps designers choose right smart home device and build a smart living space. In the design process, computer aided design skills are applied to support the stimulation of space scenario, the construction of smart home devise system, and budgeting. It is an interface for smart home device company, designer, and home user to communicate and cooperate. Cloud computing initiates the user- and service-oriented internet service; it offers multiple online platforms which promote interpersonal communication and teamwork. The designer is the bridge between the industry and user; however, designers may lack the basic understanding of smart device, which is an obstacle in the design of smart home. Currently, software supporting the smart living design is not yet available. The study suggests applying (1) the theorem of intelligent agents and 2) scenario-oriented design as the integrative framework for smart home design. The construction and stimulation on software interface helps create a smart living space, which supports the software interface in this study.
Tachibana,K., Pham,M.T., Yoshikawa,T., Furuhashi,T.
Abstract:
This note first explains Clifford's geometric algebra (GA) as a generalization of complex and quaternion algebras. Second, this note describes GA neurons as a natural extension of complex neurons. In any dimension the GA neuron takes a vector input and returns another vector output. The GA neurons are applicable to optimization of Space Folding Model for effective pattern recognition. Next, points with precision are considered using conformal geometric algebra and it is shown that addition of conformal vectors works well for precision update. The GA neuron and its use of vectors with precision (or belief) could be useful for datasets with different levels of precision/detail/belief of any dimension. The conformal vector could be also useful to set a prior distribution of geometric versors.
Aracena-Pizarro,D., Daniri-Alvarado,N.
Abstract:
This work focuses on perform 3D reconstruction from real images captured by a camera in properly controlled positions, with the aim of model an environment through successive views, in order to extract existing features on the environment, in this case, points of interest and reconstruct 3D surfaces. The main contribution of this paper consist in offering a new way to perform virtual reconstruction based on the combination of different existing vision techniques, adapting it to the needs of the environment, focusing on the SIFT method as the axis of the process, besides better integration between the epipolar geometry techniques and robust methods, and use parallel programming techniques specifically the GPU parallelism. Initial tests are performed in C + +, CUDA, TLC and VRML.
Yu,Z., Yuan,L., Luo,W.
Abstract:
Introducing Clifford Algebra as the mathematical foundation, we proposed a unified multi-dimensional GIS data model, constructed by linking data objects of different dimensions within the multivector structure of Clifford algebra. Then, algorithms for geographical network analysis (such as shortest path, minimum unicom and maximum flow analysis) and high-dimensional voronoi diagram were constructed. We use both simulation and real world data to test the usability and performance of our algorithms. The result gives very positive prospect of implement GIS analysis algorithms under Clifford Algebra framework. In that way, traditional GIS analyses algorithms can be extended not only accommodate various dimensions but also get beneficial on performance.
Getto,R., Hildenbrand,D.
Abstract:
In this communication paper we introduce our scheme for improving the estimation of curvatures in point sets presented in [Sei10a]. The basic idea of the optimization is to alter the algorithm to a more flexible point handling. The osculating circle fitting in a fixed number of directions is extended to a dynamic adjustment of the number and the orientation of the directions with regard to the chosen point. In addition, to boost the performance, the fitting itself is simplified and the fitting results are partially reused for nearby points with almost similar curvatures. Further, we use the optimized algorithm to detect shapes in the point set.
Kornev,A.
Abstract:
The diffusion process is applied to merge objects and to join discontinuities. The method combines oriented diffusion process and segmentation. The effectiveness of the method is demonstrated on a number of 3D samples. Some suggestions are made about further research and advancement of the method.
Gaitto Pereira,F.
Abstract:
In this paper we describe a method for identifying on a 3d geometry of a building, the analyses from the rendered image. The goal of this process is to link two different data types, development of a 2D analysis and link the result to the 3D model on an architectural domain.