The paper discusses implementation of advanced numerical and image-processing applications on a multiprocessor system. With a view toward 'coarse-grain' rapid prototyping, the authors implement two diverse app...
详细信息
The paper discusses implementation of advanced numerical and image-processing applications on a multiprocessor system. With a view toward 'coarse-grain' rapid prototyping, the authors implement two diverse applications onto a common framework, i.e. MAC type processors interconnected by a TESH network. TESH (Tori connected mESHes) is a recently developed interconnection network. It is hierarchical, thus allowing exploitation of computation locality as well as easy expansion, which permits efficient vLSI/ULSI realization, and appears to be well suited for 3-D vLSI/ULSI implementation. Specifically, the paper develops parallel implementation of (a) real-time solution of partial differential equations and (b) 2D wavelet transform, in such a way so as to completely hide the communication overhead. It is shown that the performance of TESH implemented algorithms is comparable to the MESH based algorithm. However, TESH networks are much easier to implement because of the significantly reduced wiring than MESH networks.
The capabilities of switched networks for parallel and distributed computers are evolving rapidly towards networks with various forms of intelligence in support of parallel execution of programs. This paper presents a...
详细信息
The capabilities of switched networks for parallel and distributed computers are evolving rapidly towards networks with various forms of intelligence in support of parallel execution of programs. This paper presents a perspective on intelligent networks, including reconfiguration of the network to adapt to the needs of successive computational algorithms being performed as part of an overall problem, for clusters containing a modest number of digital signal processors (DSPs). Scalability of the overall parallel DSP-based computer is achieved by adding these cluster nodes. It is suggested that there are many operating system functions which might be directly integrated into such an intelligent network, including adapting those operating system functions according to the needs of the specific tasks being performed. The viewpoint presented here is based on a reconfigurable system using FPGAs and being constructed for imageprocessing applications to study opportunities for integration of such intelligent networks into future silicon vLSI components, including advanced packaging such as multichip modules (MCMs). The packaging limitations present the greatest barrier to aggressive development of such networks, with active substrate MCM implementation of the network function providing the greatest flexibility and performance.
imageprocessing is usually associated with pattern recognition and is rather treated as a subject outside of the computer graphics interest. Basically computer graphics algorithms are used for the visualization of sc...
详细信息
imageprocessing is usually associated with pattern recognition and is rather treated as a subject outside of the computer graphics interest. Basically computer graphics algorithms are used for the visualization of scenes or models described using some abstract notation, while imageprocessing is used on the opposite way -i.e. when finding an abstract description of an analyzed pattern. This paper proposes to use the imageprocessing approach for rendering optical effects in computer graphics algorithms. Proposed algorithms can be used for the generation of realistic and artistic optical effects in real-time, in, for example, visual simulators, virtual reality or multi-media applications. (C) 1997 Elsevier Science B.v.
Straightforward techniques for spatial domain processing of compressed video via decompression and recompression are computationally expensive. We describe an alternative approach wherein the compressed stream is proc...
详细信息
Straightforward techniques for spatial domain processing of compressed video via decompression and recompression are computationally expensive. We describe an alternative approach wherein the compressed stream is processed in the compressed, discrete cosine transform (DCT) domain without explicit decompression and spatial domain processing, so that the output compressed stream, corresponding to the output image, conforms to the standard syntax of 8 x 8 blocks, We propose computation schemes for down-sampling and for inverse motion compensation that are applicable to any DCT-based compression method, Worst-case estimates of computation savings vary between 37% and 50% depending on the task. For typically sparse DCT blocks, the reduction in computations is more dramatic, A by-product of the proposed approach is improvement in arithmetic precision.
The paper first briefly discusses the advantages and disadvantages of using orthonormal (as opposed to non-orthogonal) bases in segmented image coding, and shows that the optimal choice is application-dependent. Next,...
详细信息
The paper first briefly discusses the advantages and disadvantages of using orthonormal (as opposed to non-orthogonal) bases in segmented image coding, and shows that the optimal choice is application-dependent. Next, it introduces fast algorithms for computing orthonormal base functions on an arbitrarily-shaped region. The algorithms are extensions of the 'natural' polynomial recursive orthogonalization (PRO) algorithm, introduced earlier by the author and differ from it in that they allow new orthogonalization orders and new types of base functions (cosines and warped polynomials in addition to ordinary polynomials). The algorithms are typically 1.5 to 3 times faster than the corresponding Gram-Schmidt (GS) methods. Three of the new algorithms, called RECT, TOTDIAG and XY are investigated in detail. The RECT and TOTDIAG algorithms are typically 15% to 30% slower than 'natural' PRO, but still 1.5 to 2.5 times faster than GS. Also, their computational advantage over GS increases;vith the number of computed base functions. A preliminary experiment shows that the combined use of the RECT or TOTDIAG base with the natural base in different areas of the image may lead to a better approximation performance, albeit at the expense of extra computations. (C) 1997 Elsevier Science B.v.
imageprocessing techniques are now considered as flexible techniques to measure and analyze various traffic parameters. Compared to the simple traffic parameters such as vehicle count and speed measurement, less atte...
详细信息
imageprocessing techniques are now considered as flexible techniques to measure and analyze various traffic parameters. Compared to the simple traffic parameters such as vehicle count and speed measurement, less attention has been paid to measure complex traffic parameters such as the Traffic Movements at Junctions (TMJ). The previous reported research on measuring traffic movements at junctions is based on full frame processing, which requires more computing power, thus is not suitable for low-cost computer systems. The method described in this paper is based on processing the key regions and detecting vehicles by using the edge detection techniques. In this approach, the architecture of the junctions is identified and the windows are placed at the key regions. A low-cost and effective edge detection technique is used to detect the vehicles and to measure the TMT parameters. The proposed algorithms have been implemented on a low-cost microcomputer system and the TMJ parameters together with other traffic parameters, such as vehicle count, are measured in real-time. (C) 1997 Elsevier Science B.v.
Automatic imageprocessingsystems suffer from high engineering costs that are necessary to adapt the configuration of the system to the problem under consideration: As a consequence, this paper presents an approach t...
详细信息
ISBN:
(纸本)0819424404
Automatic imageprocessingsystems suffer from high engineering costs that are necessary to adapt the configuration of the system to the problem under consideration: As a consequence, this paper presents an approach to reduce these costs by applying optimization methods to the setup and the configuration of an imageprocessing system. We report on experiments in which we automatically select adequate subsets of textural features from a large set of potential candidates. In addition we tell how and why training pattern selection is used as a part of the optimization process. Finally we show how genetic programming can be used for the construction of new genetic feature sets, This method has no conventional pendant and offers an interesting way to complement the algorithms made by humans.
We discuss a method for initializing the multi-wavelet decomposition algorithm by pre-filtering. The proposed pre-filtering operation projects the input signal into the space defined by the multi-scaling function asso...
详细信息
ISBN:
(纸本)0819425915
We discuss a method for initializing the multi-wavelet decomposition algorithm by pre-filtering. The proposed pre-filtering operation projects the input signal into the space defined by the multi-scaling function associated with the multi-wavelet. Since the approach is projection based, it is guaranteed to always have a solution. The space in which the original signal is contained is defined by multiple generating functions, making this work a generalization of our previous results.
We present a theory and algorithms of two dimensional multi-Gabor representations (2DMGR) using a set of nonseparable 2-D window waveforms. These windows can be customized to have different spatial and frequency orien...
详细信息
ISBN:
(纸本)0819425915
We present a theory and algorithms of two dimensional multi-Gabor representations (2DMGR) using a set of nonseparable 2-D window waveforms. These windows can be customized to have different spatial and frequency orientations, and different shapes and variances. Together, they form a frame in a 2-D spare. 2DMGRs have clear values in applications where different spatial and frequency contents and orientations of an image are important features, e.g., in texture analysis, synthesis and segmentations, and in studies of mammalian visual systems, etc.. Numerical recipe and simulation examples are also provided.
A testbed for investigation of heterogeneous and reconfigurable data network fabrics supporting a parallel DSP computational accelerator is described. The DSP processors are large-grained processors (Analog Devices SH...
详细信息
A testbed for investigation of heterogeneous and reconfigurable data network fabrics supporting a parallel DSP computational accelerator is described. The DSP processors are large-grained processors (Analog Devices SHARC DSPs), with a variety of parallel DSP array architectures possible. The network fabric is intended to be reconfigurable (within a rich but necessarily limited set of structures) to adapt to the needs of a sequence of imageprocessingalgorithms being executed (e.g., in a medical imageprocessing environment). The testbed will exploit conventional FPGA components to provide reconfigurable network structures and will exploit commercial high-speed interconnect components emerging for applications such as board-to-board applications. As a computational accelerator, the testbed is intended to be controlled by a host processor, with the host processor cooperating in the definition of the changes in the structure of the network structure as execution of a sequence of imageprocessingalgorithms proceeds.
暂无评论