Efficient image filtering and information reduction in reconfigurable logic

Tørresen, Jim; Bakke, J.W.; Sekanina, Lukáš

doi:10.1109/norchp.2004.1423823

Cited by 4 publications

(3 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Massive data applications, as image analyses or high precise numeric model simulations, in general, require a high performance computation approach [1] [3][4] [10]. Therefore, architectures to face such kind of problems should be able to provide such resources and its intrinsic parallelism.…”

Section: Introductionmentioning

confidence: 99%

Mapping of Massive Data Processing Systems to FPGA Computers Based on Temporal Partitioning and Design Space Exploration

Nascimento¹,

Silva²,

Seixas³

et al. 2020

JICS

View full text Add to dashboard Cite

High parallelism degree is fundamental for high speed massive data processing systems. Modern FPGA devices can provide such parallelism plus flexibility. However, these devices are still limited by their logic block size, memory size, memory bandwidth and configuration time. Temporal partitioning techniques can be a solution for such problems when FPGAs are used to implement large systems. In this case, the system is split into partitions (called contexts), multiplexed in a FPGA, by using reconfiguration techniques. This approach can increase the effective area for system implementation, allowing increase of parallelism in each task that composes the application. However, the necessary reconfiguration time between contexts can cause performance decrease. A possible solution for this is an intensive parallelism exploration of massive data application to compensate for this overhead and improve global performance. This is true for modern FPGA with relatively high reconfiguration speed. In this work, A reconfigurable computer platform and design space exploration techniques are proposed for mapping of such massive data applications, as image processing, in FPGA devices, depending on the application task scheduling. A library with different hardware implementation for a different parallelism degree is used for better adjustment of space/time for each task. Experiments demonstrate the efficiency of this approach when compared to the optimal mapping reached by exhaustive timing search in the complete design space exploration. A design flow is shown based on library components that implements typical tasks used in the domain of applications.

show abstract

Section: Introductionmentioning

confidence: 99%

Mapping of Massive Data Processing Systems to FPGA Computers Based on Temporal Partitioning and Design Space Exploration

Nascimento¹,

Silva²,

Seixas³

et al. 2020

JICS

View full text Add to dashboard Cite

show abstract

“…There are other methods that use reconfigurable hardware 12 or processor adapted software libraries such as Intel's IPL 13 to accelerate the filtering process. GPU implementations of image filtering methods became also popular, especially for the computation of area sums using pyramid representations.…”

Section: Introductionmentioning

confidence: 99%

Reshuffling: a fast algorithm for filtering with arbitrary kernels

Porikli

2008

Real-Time Image Processing 2008

View full text Add to dashboard Cite

A Novel method to accelerate the application of linear filters that have multiple identical coefficients on arbitrary kernels is presented. Such filters, including Gabor filters, gray level morphological operators, volume smoothing functions, etc., are wide used in many computer vision tasks. By taking advantage of the overlapping area between the kernels of the neighboring points, the reshuffling technique prevents from the redundant multiplications when the filter response is computed. It finds a set of unique, constructs a set of relative links for each coefficient, and then sweeps through the input data by accumulating the responses at each point while applying the coefficients using their relative links. Dual solutions, single input access and single output access, that achieve 40% performance improvement are provided. In addition to computational advantage, this method keeps a minimal memory imprint, which makes it an ideal method for embedded platforms. The effects of quantization, kernel size, and symmetry on the computational savings are discussed. Results prove that the reshuffling is superior to the conventional approach. Real-Time Image Proc., Feb. 2008 This work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonprofit educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Research Laboratories, Inc.; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Research Laboratories, Inc. All rights reserved. SPIE Electronic Imaging, Proceedings of ABSTRACTA novel method to accelerate the application of linear filters that have multiple identical coefficients on arbitrary kernels is presented. Such filters, including Gabor filters, gray level morphological operators, volume smoothing functions, etc., are widely used in many computer vision tasks. By taking advantage of the overlapping area between the kernels of the neighboring points, the reshuffling technique prevents from the redundant multiplications when the filter response is computed. It finds a set of unique coefficients, constructs a set of relative links for each coefficient, and then sweeps through the input data by accumulating the responses at each point while applying the coefficients using their relative links. Dual solutions, single input access and single output access, that achieve 40% performance improvement are provided. In addition to computational advantage, this method keeps a minimal memory imprint, which makes it an ideal method for embedded platforms. The effects of quantization, kernel size, and symmetry on the computational savings are discussed.Results prove that the reshuffling is superior to the...

show abstract

“…Hardware implementation results in terms of visual performance, speed, and area consumption of the implementation is given, yet the algorithm is application dependent and requires some training period for perfect segmentation. Jim et al [16] used an experimental colour representation for filtering common colours found in road signs. They used RGB values as opposed to HSV to avoid computationally expensive conversions as outputs of many cameras are in RGB mode.…”

Section: Introductionmentioning

confidence: 99%