Membrane computing (known as P systems) is a novel class of distributedparallel computing models inspired by the structure and functioning of living cells and organs, and its application to the real-world problems ha...
详细信息
Membrane computing (known as P systems) is a novel class of distributedparallel computing models inspired by the structure and functioning of living cells and organs, and its application to the real-world problems has become a hot topic in recent years. This paper discusses an interesting open problem in digital watermarking domain, optimal watermarking problem, and proposes a new optimal image watermarking method under the framework of P systems. A special membrane structure is designed and its cells as parallel computing units are used to find the optimal watermarking parameters for image blocks. Some cells use the position-velocity model to evolve watermarking parameters of image blocks, while another cell evaluates the objects in the system. In addition to the evolution rules, communication rules are used to exchange and share information between the cells. Simulation experiments on large image set compare the proposed framework with other existing watermarking methods and demonstrate its superiority. (C) 2014 Elsevier B.V. All rights reserved.
Deep learning, recently, has been successfully applied to image classification, object recognition and speech recognition. However, the benefits of deep learning and accompanying architectures have been largely unknow...
详细信息
ISBN:
(纸本)9781479988518
Deep learning, recently, has been successfully applied to image classification, object recognition and speech recognition. However, the benefits of deep learning and accompanying architectures have been largely unknown for BCI applications. In motor imagery-based BCI, an energy-based feature, typically after spatial filtering, is commonly used for classification. Although this feature corresponds to the estimate of event-related synchronization/desynchronization in the brain, it neglects energy dynamics which may contain valuable discriminative information. Because traditional classification methods, such as SVM, cannot handle this dynamical property, we proposed an architecture that inputs a dynamic energy representation of EEG data and utilizes convolutional neural networks for classification. By combining this network with a static energy network, we saw a significant increase in performance. We evaluated the proposed method and compared with SVM on a multi-class motor imagery dataset (BCI competition dataset iv-2a). Our method outperforms SVM with static energy features significantly (p < 0.01).
Current computer forensics tools have some limitations on anti-forensics attacks, cloud computing, and a large increase in the size of forensics targets. To solve these problems, this paper proposes a system that pres...
详细信息
Current computer forensics tools have some limitations on anti-forensics attacks, cloud computing, and a large increase in the size of forensics targets. To solve these problems, this paper proposes a system that preserves storage data on virtual machines by acquiring all data sectors with time stamps. The proposed system can restore a previous state of a block device at any date and time that is specified by an investigator. The proposed system aims to monitor users' behavior in Infrastructure-as-a-Service (IaaS) cloud platforms. This paper also presents a rapid file detection system that finds a target file from a large collection of the acquired data sectors by using sector-hashes and paralleldistributedprocessing. This system enables investigators to track and to find a target file that is related to incidents or crimes in the cloud. First, this paper reports the preliminary experiments of a sector-hash based file detection method on three major operating systems for evaluating its effectiveness. We present a design and an implementation of the proposed monitoring and target file detection system by using Xen hypervisor and MapReduce. We report results of its performance evaluation. Finally, we discuss possible methods to improve the performance and the limitations of the current proposed mechanism.
In this paper, we investigate the use of random selection (RS) and random projection (RP) for hyperspectral image analysis, which are data-independent and computationally more efficient than other widely used dimensio...
详细信息
ISBN:
(纸本)9781628413106
In this paper, we investigate the use of random selection (RS) and random projection (RP) for hyperspectral image analysis, which are data-independent and computationally more efficient than other widely used dimensionality reduction methods. Both anomaly detection and target detection are considered. Due to the random nature, multiple runs of RS or RP are conducted followed by decision fusion to ensure a stable output. parallel implementations using graphics processing unit (GPU) and clusters are also investigated. The experimental results demonstrated that both RS and RP are capable of providing better target detection performance after decision fusion, while the overall computing time can be greatly decreased with parallel implementations.
Computed Tomography (CT) scanners evolved from simple parallel-beam geometry into more complex fan-beam geometry. The rebinning mechanism to convert fan-beam projections to parallel-beam projections is one of the meth...
详细信息
ISBN:
(纸本)9781479944354
Computed Tomography (CT) scanners evolved from simple parallel-beam geometry into more complex fan-beam geometry. The rebinning mechanism to convert fan-beam projections to parallel-beam projections is one of the methods simplifying the reconstruction of the CT image. Various interpolation methods result in different numerical presentations and noisy textures in the reconstructed CT images. This paper evaluates three interpolation methods, linear, frequency domain and higher order in rebinning fan-beam data into parallel-beam data in the framework of the FBP algorithm in parallel-beam CTs. The high-contrast spatial resolution, modulation transfer function (MTF), and noise are used to quantify the image quality. The noise is expressed as the noise power spectrum (NPS) and noise variance of the reconstructed image containing only Poisson distributed noise. Our results show that the higher order interpolation, cubic-spline, is the most helpful to improve the image quality in rebinning. We further demonstrate that the rebinning data gives more satisfactory results for assessment of noise variance in comparison with the fan-beam data.
High Performance Computing (HPC) systems are nowadays more and more heterogeneous. Different processor types can be found on a single node including accelerators such as Graphics processing Units (GPUs). To cope with ...
详细信息
High Performance Computing (HPC) systems are nowadays more and more heterogeneous. Different processor types can be found on a single node including accelerators such as Graphics processing Units (GPUs). To cope with the challenge of programming such complex systems, this work presents a domain-specific approach to automatically generate code tailored to different processor types. Low-level CUDA and OpenCL code is generated from a high-level description of an algorithm specified in a Domain-Specific Language (DSL) instead of writing hand-tuned code for GPU accelerators. The DSL is part of the Heterogeneous imageprocessing Acceleration (HIPA(cc)) framework and was extended in this work to handle grid hierarchies in order to model different cycle types. Language constructs are introduced to process and represent data at different resolutions. This allows to describe imageprocessing algorithms that work on image pyramids as well as multigrid methods in the stencil domain. By decoupling the algorithm from its schedule, the proposed approach allows to generate efficient stencil code implementations. Our results show that similar performance compared to hand-tuned codes can be achieved. (C) 2014 Elsevier Inc. All rights reserved.
This paper reviews and compares the performance of several methods to detect target tracks in image sequences. The targets are assumed to be sub-pixel or not resolved by the imaging system, and moving over a static ba...
详细信息
ISBN:
(纸本)9781628410266
This paper reviews and compares the performance of several methods to detect target tracks in image sequences. The targets are assumed to be sub-pixel or not resolved by the imaging system, and moving over a static background. To process the resulting large amount of data requires simple, fast and robust processingmethods to quickly find and display tracks of moving targets in a single image. An object moving through a pixel in a scene will momentarily perturb the pixel intensity signal, introducing a change of both skewness and kurtosis in the intensity histogram relative to an undisturbed pixel. Numerical experiments show that for Gaussian and Poisson distributed system noise higher order moments (>2) perform better than second order detectors.
In computer graphics, generating high-quality images at high frame rates for render- ing complex scenes is a challenging task. A well-known approach to tackling this important task is to utilize parallelprocessing th...
In computer graphics, generating high-quality images at high frame rates for render- ing complex scenes is a challenging task. A well-known approach to tackling this important task is to utilize parallelprocessing through distributing rendering and sim- ulation tasks to different processing units. In this thesis, several methods of distributed rendering architectures are investigated, and the bottlenecks in distributed rendering are analyzed. Based on this analysis, guidelines for distributed rendering in a network of computers are proposed. Moreover, in the thesis, an efficient load balancing strategy is proposed for distribut- ing the rendering of individual frames to different processing units in a network. In this distributed rendering heterogeneous system, there are computers equipped with multiple Graphical processing Units (GPUs) with different rendering performances all in the same network with a server, which collects rendering performances of the GPUs in the different image Generators (IGs) based on an effective load balancing. By means of the novel load balancing strategy, the thesis shows that such a system can increase the rendering performance of slow computers with the help of the fast ones. Lastly, this model is extended to develop an adaptive hybrid model where (i) parts of a frame or a scene can be distributed and (ii) GPU-GPU and GPU-CPU distributions can be considered. This model can adjust itself to the changing loads of the GPUs and determine an efficient load balancing strategy for distributed rendering.
This paper presents a parallel remeshing algorithm for distributed-memory architectures. It is an iterative parallel algorithm that divides the areas to be remeshed into multiple pieces which can be distributed to as ...
详细信息
ISBN:
(纸本)9788494284472
This paper presents a parallel remeshing algorithm for distributed-memory architectures. It is an iterative parallel algorithm that divides the areas to be remeshed into multiple pieces which can be distributed to as many processing elements as possible, in order for these pieces to be remeshed concurrently by a third-party sequential remesher. Then, remeshed pieces are reintegrated into the distributed mesh, and this process is iterated until all relevant areas of the mesh have been remeshed. Any sequential remesher can be used, provided it allows some of the mesh elements not to be modified, so as to preserve interfaces between pieces. Our method, which has been implemented in the PaMPA library, is validated by a set of experiments involving both isotropic and anisotropic meshes.
This paper presents a distributed solution for the development of deformable model-based medical image segmentation methods. The design and implementation stages of the segmentation methods usually require a lot of ti...
详细信息
暂无评论