Fast bilateral filtering using recursive moving sum

• Igarashi Masaki

  Graduate School of Information Science and Technology, Hokkaido University

• Ikebe Masayuki

  Graduate School of Information Science and Technology, Hokkaido University

• Shimoyama Sohsuke

  Graduate School of Information Science and Technology, Hokkaido University

• Motohisa Junichi

  Graduate School of Information Science and Technology, Hokkaido University

Abstract

We propose a constant-time algorithm for a bilateral filter. Bilateral filter can be converted into the operation of three-dimensional (3D) convolution. By using recursive moving sum, we can reduce the number of calculations needed to construct a pseudo-Gaussian filter. Applying one-dimensional Gaussian filter to the 3D convolution, we achieved a constant-time bilateral filter. We used a 3-GHz CPU without SIMD instructions, or multi-thread operations. We confirmed our proposed bilateral filter to be processed in constant time. In practical conditions, high PSNR values over 40 dB are obtained.

Journal

• Nonlinear Theory and Its Applications, IEICE

  Nonlinear Theory and Its Applications, IEICE 3 (2), 222-232, 2012

  The Institute of Electronics, Information and Communication Engineers

Keywords

• image processing
• bilateral filter
• constant-time algorithm

Details

• CRID: 1390282680323167488
• NII Article ID: 130002139293
• DOI: 10.1587/nolta.3.222
• BIBCODE: 2012NTA.....3..222I
• ISSN: 21854106
• Web Site: https://www.jstage.jst.go.jp/article/nolta/3/2/3_2_222/_pdf
• Text Lang: en
• Data Source:
  • JaLC
  • Crossref
  • CiNii Articles
  • KAKEN
• Abstract License Flag: Disallowed