In this paper we introduce the Reduction Sweep algorithm, a novel graph-based image segmentation algorithm that is designed for easy parallelization. It is based on a clustering approach focusing on local image charac...
详细信息
In this paper we introduce the Reduction Sweep algorithm, a novel graph-based image segmentation algorithm that is designed for easy parallelization. It is based on a clustering approach focusing on local image characteristics. Each pixel is compared with its neighbors in an implicitly independent manner, and those deemed sufficiently similar according to a color criterion are joined. We achieve fast execution times while still maintaining the visual quality of the results. The algorithm is presented in four different implementations: sequential CPU, parallel CPU, GPU, and hybrid CPU-GPU. We compare the execution times of the four versions with each other and with other closely related image segmentation algorithms.
Barcode technology is essential in automatic identification, and is used in a wide range of real-time applications. Different code types and applications impose special problems, so there is a continuous need for solu...
详细信息
This work introduces a residual operator called ultimate attribute leveling. We also present an efficient algorithm for ultimate attribute leveling computation by using a structure called tree of shapes. Our algorithm...
详细信息
This work introduces a residual operator called ultimate attribute leveling. We also present an efficient algorithm for ultimate attribute leveling computation by using a structure called tree of shapes. Our algorithm for computating ultimate attribute leveling is based on the fact (proved in this work) that levelings can be obtained by pruning nodes from the tree of shapes. This is a novel result, since so far it is known that levelings can be obtained from component trees. Finally, we propose the use of ultimate attribute leveling with shape information to extract contrast using a priori knowledge of an application. Experimental results applied to text location show the potentiality of using ultimate attribute leveling with shape information for solving problems in imageprocessing area.
Several methods have been proposed to overcome the pixel art scaling problem through the years. In this article we describe a novel approach to be applied through a massively parallel architecture that can address thi...
详细信息
Several methods have been proposed to overcome the pixel art scaling problem through the years. In this article we describe a novel approach to be applied through a massively parallel architecture that can address this issue in real time. To achieve this we design a local and context independent algorithm that enables an efficient parallel implementation on the GPU, delivering full frames output at response time for the user interaction. Our main goal is to apply the method on full frames of old games, which were based on pixel art graphics until the half of the 1990's, and keep the output frame rate good enough for playing.
Accelerators are an increasingly common option to boost performance of codes that require extensive number crunching. In this paper we report on our experience with NVIDIA accelerators to study fluid systems using the...
详细信息
ISBN:
(纸本)9781479929276
Accelerators are an increasingly common option to boost performance of codes that require extensive number crunching. In this paper we report on our experience with NVIDIA accelerators to study fluid systems using the Lattice Boltzmann (LB) method. The regular structure of LB algorithms makes them suitable for processor architectures with a large degree of parallelism, such as recent multi-and many-core processors and GPUs;however, the challenge of exploiting a large fraction of the theoretically available performance of this new class of processors is not easily met. We consider a state-of-the-art two-dimensional LB model based on 37 populations (a D2Q37 model), that accurately reproduces the thermo-hydrodynamics of a 2D-fluid obeying the equation-of-state of a perfect gas. The computational features of this model make it a significant benchmark to analyze the performance of new computational platforms, since critical kernels in this code require both high memory-bandwidth on sparse memory addressing patterns and floating-point throughput. In this paper we consider two recent classes of GPU boards based on the Fermi and Kepler architectures;we describe in details all steps done to implement and optimize our LB code and analyze its performance first on single-GPU systems, and then on parallel multi-GPU systems based on one node as well as on a cluster of many nodes;in the latter case we use CUDA-aware MPI as an abstraction layer to assess the advantages of advanced GPU-to-GPU communication technologies like GPUDirect. On our implementation, aggregate sustained performance of the most compute intensive part of the code breaks the 1 double-precision Tflops barrier on a single-host system with two GPUs.
There are three possible partitionings of the continuous plane into regular polygons that leads to triangular, square, and hexagonal grids. The topology of the square grid is fairly well-understood, but it cannot be s...
详细信息
There are three possible partitionings of the continuous plane into regular polygons that leads to triangular, square, and hexagonal grids. The topology of the square grid is fairly well-understood, but it cannot be said of the remaining two regular sampling schemes. This paper presents a general characterization of simple pixels and some simplified sufficient conditions for topology-preserving operators in all the three types of regular grids.
With the advances in central processing unit (CPU) capability, 3D display technology becomes popular in recent years. Now, there are more and more 3D products such as 3D camera, 3D projector, and 3D-TV. The 3D technol...
详细信息
ISBN:
(纸本)9781467357623;9781467357609
With the advances in central processing unit (CPU) capability, 3D display technology becomes popular in recent years. Now, there are more and more 3D products such as 3D camera, 3D projector, and 3D-TV. The 3D technology is not difficult to understand because most of the video contents are captured through two individual lenses. The corresponding video contents are considered as two respective bit streams. However, how to transform traditional 2D video contents to 3D one is a critical problem to solve it. This paper proposes a real-time 3D rendering architecture for view synthesis in 2D-to-3D conversion. The real-time architecture can support 60 frames per second and full HD resolution (1920x1080) on FPGA platform.
Modeling organic objects and natural phenomena is a very time consuming task, and usually requires a certain level of expertise. This happens due to the lack of suitable geometric primitives available in most modeling...
详细信息
Modeling organic objects and natural phenomena is a very time consuming task, and usually requires a certain level of expertise. This happens due to the lack of suitable geometric primitives available in most modeling tools. In addition, with the great advance in real-time rendering power, there is a growing demand for more complex models. Hence, this paper proposes a method to automatically build visually convincing spider orb-webs that are self-adapted to the environment. We show results of mesh-based spider orb-webs built in real-time that are visually pleasing and realistically looking.
Object tracking in video sequences is still a challenging issue in real-time video processing. In this paper we propose an integrated detection and tracking method suitable for high-definition videos at real-time fram...
详细信息
Object tracking in video sequences is still a challenging issue in real-time video processing. In this paper we propose an integrated detection and tracking method suitable for high-definition videos at real-time frame rates. In this method we implement a frame segmentation procedure using integral images of the background, which permits to discard the analysis of several image parts of each frame and achieve high frame rates. Also we extend the proposed algorithm to detect multiple objects in parallel.
Vector field analysis and visualization is a fundamental tool in science and engineering applications, raising the need for robust methods to capture and represent the global vector field behavior. One such representa...
详细信息
Vector field analysis and visualization is a fundamental tool in science and engineering applications, raising the need for robust methods to capture and represent the global vector field behavior. One such representation, the topological graph, partitions the domain into regions where the flow of the vector field is homogeneous. This work introduces a topological graph construction based on streamlines. It is guaranteed to produce a coherent result even when some singularities are not detected. This work also details an application of topological graphs to improve the generation of self-animated images. In this application, the streamline-based approach carries almost no overhead, since self-animated images already rely on streamlines, but leads to a threefold speed-up of the core processing.
暂无评论