Multigrid Bilateral Filtering

Qingyun SHE; Zongqing LU; Weifeng LI; Qingmin LIAO

doi:10.1587/transinf.2014EDP7070

Multigrid Bilateral Filtering

Qingyun SHE, Zongqing LU, Weifeng LI, Qingmin LIAO

Full Text Views

0

Cite this

Summary :

The bilateral filter (BF) is a nonlinear and low-pass filter which can smooth an image while preserving detail structures. However, the filer is time consuming for real-time processing. In this paper, we bring forward a fresh idea that bilateral filtering can be accelerated by a multigrid (MG) scheme. Our method is based on the following two facts. a) The filtering result by a BF with a large kernel size on the original resolution can be approximated by applying a small kernel sized (3×3) version on the lower resolution many times on the premise of visual acceptance. Early work has shown that a BF can be viewed as nonlinear diffusion. The desired filtering result is actually an intermediate status of the diffusion process. b) Iterative linear equation techniques are sufficiently mature to cope with the nonlinear diffusion equation, which can be accelerated by the MG scheme. Experimental results with both simulated data sets and real sets are provided, and the new method is demonstrated to achieve almost twice the speed of the state-of-the-art. Compared with previous efforts for finding a generalized representation to link bilateral filtering and nonlinear diffusion by adaptive filtering, a novel relationship between nonlinear diffusion and bilateral filtering is explored in this study by focusing attention on numerical calculus.

Publication: IEICE TRANSACTIONS on Information Vol.E97-D No.10 pp.2748-2759

Publication Date: 2014/10/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1587/transinf.2014EDP7070

Type of Manuscript: PAPER

Category: Image Processing and Video Processing

Authors

Qingyun SHE
  Tsinghua University
Zongqing LU
  Tsinghua University
Weifeng LI
  Tsinghua University
Qingmin LIAO
  Tsinghua University

Keyword

bilateral filter, multigrid, nonlinear diffusion

Cite this

Copy

Qingyun SHE, Zongqing LU, Weifeng LI, Qingmin LIAO, "Multigrid Bilateral Filtering" in IEICE TRANSACTIONS on Information, vol. E97-D, no. 10, pp. 2748-2759, October 2014, doi: 10.1587/transinf.2014EDP7070.
Abstract: The bilateral filter (BF) is a nonlinear and low-pass filter which can smooth an image while preserving detail structures. However, the filer is time consuming for real-time processing. In this paper, we bring forward a fresh idea that bilateral filtering can be accelerated by a multigrid (MG) scheme. Our method is based on the following two facts. a) The filtering result by a BF with a large kernel size on the original resolution can be approximated by applying a small kernel sized (3×3) version on the lower resolution many times on the premise of visual acceptance. Early work has shown that a BF can be viewed as nonlinear diffusion. The desired filtering result is actually an intermediate status of the diffusion process. b) Iterative linear equation techniques are sufficiently mature to cope with the nonlinear diffusion equation, which can be accelerated by the MG scheme. Experimental results with both simulated data sets and real sets are provided, and the new method is demonstrated to achieve almost twice the speed of the state-of-the-art. Compared with previous efforts for finding a generalized representation to link bilateral filtering and nonlinear diffusion by adaptive filtering, a novel relationship between nonlinear diffusion and bilateral filtering is explored in this study by focusing attention on numerical calculus.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2014EDP7070/_p

Copy

@ARTICLE{e97-d_10_2748,
author={Qingyun SHE, Zongqing LU, Weifeng LI, Qingmin LIAO, },
journal={IEICE TRANSACTIONS on Information},
title={Multigrid Bilateral Filtering},
year={2014},
volume={E97-D},
number={10},
pages={2748-2759},
abstract={The bilateral filter (BF) is a nonlinear and low-pass filter which can smooth an image while preserving detail structures. However, the filer is time consuming for real-time processing. In this paper, we bring forward a fresh idea that bilateral filtering can be accelerated by a multigrid (MG) scheme. Our method is based on the following two facts. a) The filtering result by a BF with a large kernel size on the original resolution can be approximated by applying a small kernel sized (3×3) version on the lower resolution many times on the premise of visual acceptance. Early work has shown that a BF can be viewed as nonlinear diffusion. The desired filtering result is actually an intermediate status of the diffusion process. b) Iterative linear equation techniques are sufficiently mature to cope with the nonlinear diffusion equation, which can be accelerated by the MG scheme. Experimental results with both simulated data sets and real sets are provided, and the new method is demonstrated to achieve almost twice the speed of the state-of-the-art. Compared with previous efforts for finding a generalized representation to link bilateral filtering and nonlinear diffusion by adaptive filtering, a novel relationship between nonlinear diffusion and bilateral filtering is explored in this study by focusing attention on numerical calculus.},
keywords={},
doi={10.1587/transinf.2014EDP7070},
ISSN={1745-1361},
month={October},}

Copy

TY - JOUR
TI - Multigrid Bilateral Filtering
T2 - IEICE TRANSACTIONS on Information
SP - 2748
EP - 2759
AU - Qingyun SHE
AU - Zongqing LU
AU - Weifeng LI
AU - Qingmin LIAO
PY - 2014
DO - 10.1587/transinf.2014EDP7070
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E97-D
IS - 10
JA - IEICE TRANSACTIONS on Information
Y1 - October 2014
AB - The bilateral filter (BF) is a nonlinear and low-pass filter which can smooth an image while preserving detail structures. However, the filer is time consuming for real-time processing. In this paper, we bring forward a fresh idea that bilateral filtering can be accelerated by a multigrid (MG) scheme. Our method is based on the following two facts. a) The filtering result by a BF with a large kernel size on the original resolution can be approximated by applying a small kernel sized (3×3) version on the lower resolution many times on the premise of visual acceptance. Early work has shown that a BF can be viewed as nonlinear diffusion. The desired filtering result is actually an intermediate status of the diffusion process. b) Iterative linear equation techniques are sufficiently mature to cope with the nonlinear diffusion equation, which can be accelerated by the MG scheme. Experimental results with both simulated data sets and real sets are provided, and the new method is demonstrated to achieve almost twice the speed of the state-of-the-art. Compared with previous efforts for finding a generalized representation to link bilateral filtering and nonlinear diffusion by adaptive filtering, a novel relationship between nonlinear diffusion and bilateral filtering is explored in this study by focusing attention on numerical calculus.
ER -