Fast image filtering by DCT-based kernel decomposition and sequential sum update

Research output: Chapter in Book/Report/Conference proceedingConference contribution

5 Citations (Scopus)

Abstract

This paper presents an approximate Gaussian filter which can run in one-pass with high accuracy based on spectrum sparsity. This method is a modification of the cosine integral image (CII), which decomposes a filter kernel into few cosine terms and convolves each cosine term with an input image in constant time per pixel by using integral images and look-up tables. However, they require much workspace and high access cost. The proposed method solves the problem with no decline in quality by sequentially updating sums instead of integral images and by improving look-up tables, which accomplishes a one-pass approximation with much less workspace. A specialization for tiny kernels are also discussed for faster calculation. Experiments on image filtering show that the proposed method can run nearly two times faster than CII and also than convolution even with small kernel.

Original languageEnglish
Title of host publicationProceedings - International Conference on Image Processing, ICIP
Pages125-128
Number of pages4
DOIs
Publication statusPublished - 2012
Event2012 19th IEEE International Conference on Image Processing, ICIP 2012 - Lake Buena Vista, FL
Duration: 2012 Sep 302012 Oct 3

Other

Other2012 19th IEEE International Conference on Image Processing, ICIP 2012
CityLake Buena Vista, FL
Period12/9/3012/10/3

Fingerprint

Convolution
Pixels
Decomposition
Costs
Experiments

Keywords

  • digital signal processing
  • discrete cosine transform
  • Gaussian filter
  • sparse spectrum

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems

Cite this

Sugimoto, K., & Kamata, S. (2012). Fast image filtering by DCT-based kernel decomposition and sequential sum update. In Proceedings - International Conference on Image Processing, ICIP (pp. 125-128). [6466811] https://doi.org/10.1109/ICIP.2012.6466811

Fast image filtering by DCT-based kernel decomposition and sequential sum update. / Sugimoto, Kenjiro; Kamata, Seiichiro.

Proceedings - International Conference on Image Processing, ICIP. 2012. p. 125-128 6466811.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Sugimoto, K & Kamata, S 2012, Fast image filtering by DCT-based kernel decomposition and sequential sum update. in Proceedings - International Conference on Image Processing, ICIP., 6466811, pp. 125-128, 2012 19th IEEE International Conference on Image Processing, ICIP 2012, Lake Buena Vista, FL, 12/9/30. https://doi.org/10.1109/ICIP.2012.6466811
Sugimoto K, Kamata S. Fast image filtering by DCT-based kernel decomposition and sequential sum update. In Proceedings - International Conference on Image Processing, ICIP. 2012. p. 125-128. 6466811 https://doi.org/10.1109/ICIP.2012.6466811
Sugimoto, Kenjiro ; Kamata, Seiichiro. / Fast image filtering by DCT-based kernel decomposition and sequential sum update. Proceedings - International Conference on Image Processing, ICIP. 2012. pp. 125-128
@inproceedings{95c6bbda58124b6a8aa1e06eb32634ca,
title = "Fast image filtering by DCT-based kernel decomposition and sequential sum update",
abstract = "This paper presents an approximate Gaussian filter which can run in one-pass with high accuracy based on spectrum sparsity. This method is a modification of the cosine integral image (CII), which decomposes a filter kernel into few cosine terms and convolves each cosine term with an input image in constant time per pixel by using integral images and look-up tables. However, they require much workspace and high access cost. The proposed method solves the problem with no decline in quality by sequentially updating sums instead of integral images and by improving look-up tables, which accomplishes a one-pass approximation with much less workspace. A specialization for tiny kernels are also discussed for faster calculation. Experiments on image filtering show that the proposed method can run nearly two times faster than CII and also than convolution even with small kernel.",
keywords = "digital signal processing, discrete cosine transform, Gaussian filter, sparse spectrum",
author = "Kenjiro Sugimoto and Seiichiro Kamata",
year = "2012",
doi = "10.1109/ICIP.2012.6466811",
language = "English",
isbn = "9781467325332",
pages = "125--128",
booktitle = "Proceedings - International Conference on Image Processing, ICIP",

}

TY - GEN

T1 - Fast image filtering by DCT-based kernel decomposition and sequential sum update

AU - Sugimoto, Kenjiro

AU - Kamata, Seiichiro

PY - 2012

Y1 - 2012

N2 - This paper presents an approximate Gaussian filter which can run in one-pass with high accuracy based on spectrum sparsity. This method is a modification of the cosine integral image (CII), which decomposes a filter kernel into few cosine terms and convolves each cosine term with an input image in constant time per pixel by using integral images and look-up tables. However, they require much workspace and high access cost. The proposed method solves the problem with no decline in quality by sequentially updating sums instead of integral images and by improving look-up tables, which accomplishes a one-pass approximation with much less workspace. A specialization for tiny kernels are also discussed for faster calculation. Experiments on image filtering show that the proposed method can run nearly two times faster than CII and also than convolution even with small kernel.

AB - This paper presents an approximate Gaussian filter which can run in one-pass with high accuracy based on spectrum sparsity. This method is a modification of the cosine integral image (CII), which decomposes a filter kernel into few cosine terms and convolves each cosine term with an input image in constant time per pixel by using integral images and look-up tables. However, they require much workspace and high access cost. The proposed method solves the problem with no decline in quality by sequentially updating sums instead of integral images and by improving look-up tables, which accomplishes a one-pass approximation with much less workspace. A specialization for tiny kernels are also discussed for faster calculation. Experiments on image filtering show that the proposed method can run nearly two times faster than CII and also than convolution even with small kernel.

KW - digital signal processing

KW - discrete cosine transform

KW - Gaussian filter

KW - sparse spectrum

UR - http://www.scopus.com/inward/record.url?scp=84875826067&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84875826067&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2012.6466811

DO - 10.1109/ICIP.2012.6466811

M3 - Conference contribution

AN - SCOPUS:84875826067

SN - 9781467325332

SP - 125

EP - 128

BT - Proceedings - International Conference on Image Processing, ICIP

ER -