Detail publikace

Parallel Implementation of 2D Forward Discrete Wavelet Transform on multicore CPUs

Originální název

Parallel Implementation of 2D Forward Discrete Wavelet Transform on multicore CPUs

Anglický název

Parallel Implementation of 2D Forward Discrete Wavelet Transform on multicore CPUs

Jazyk

en

Originální abstrakt

In this paper we present parallel implementation of 2D forward DTWT using C++ and Intel Threading Building Blocks library. Developed algorithm is described in detail and reached speedup is shown. The image is processed blockwise with certain overlap between blocks so that the result it equal to the result obtained by processing the image as a whole. Consequently, the blocking artifact is completely eliminated. The exact dimensions of the overlap is given by the developed algorithm. The algorithm works with any block size also the blocks can have different dimensions. The overlap given by the algorithm ensures that blocks can be processed independently. Necessary redundancy is discussed also.

Anglický abstrakt

In this paper we present parallel implementation of 2D forward DTWT using C++ and Intel Threading Building Blocks library. Developed algorithm is described in detail and reached speedup is shown. The image is processed blockwise with certain overlap between blocks so that the result it equal to the result obtained by processing the image as a whole. Consequently, the blocking artifact is completely eliminated. The exact dimensions of the overlap is given by the developed algorithm. The algorithm works with any block size also the blocks can have different dimensions. The overlap given by the algorithm ensures that blocks can be processed independently. Necessary redundancy is discussed also.

Dokumenty

BibTex


@article{BUT50666,
  author="Zdeněk {Průša} and Pavel {Rajmic}",
  title="Parallel Implementation of 2D Forward Discrete Wavelet Transform on multicore CPUs",
  annote="In this paper we present parallel implementation of 2D forward DTWT using C++ and Intel Threading Building Blocks library. Developed algorithm is described in detail and reached speedup is shown. The image is processed blockwise with certain overlap between blocks so that the result it equal to the result obtained by processing the image as a whole. Consequently, the blocking artifact is completely eliminated. The exact dimensions of the overlap is given by the developed algorithm. The algorithm works with any block size also the blocks can have different dimensions. The overlap given by the algorithm ensures that blocks can be processed independently. Necessary redundancy is discussed also.",
  chapter="50666",
  journal="Elektrorevue - Internetový časopis (http://www.elektrorevue.cz)",
  number="112",
  volume="2010",
  year="2010",
  month="november",
  pages="1--7",
  type="journal article - other"
}