Image denoising using mixtures of gaussian scale mixtures

Jose A. Guerrero-Colón, Eero Simoncelli, Javier Portilla

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

Abstract

The local statistical properties of photographic images, when represented in a multi-scale basis, have been described using Gaussian scale mixtures (GSMs). In that model, each spatial neighborhood of coefficients is described as a Gaussian random vector modulated by a random hidden positive scaling variable. Here, we introduce a more powerful model in which neighborhoods of each subband are described as a finite mixture of GSMs. We develop methods to learn the mixing densities and covariance matrices associated with each of the GSM components from a single image, and show that this process naturally segments the image into regions of similar content. The model parameters can also be learned in the presence of additive Gaussian noise, and the resulting fitted model may be used as a prior for Bayesian noise removal. Simulations demonstrate this model substantially outperforms the original GSM model.

Original languageEnglish (US)
Title of host publication2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings
Pages565-568
Number of pages4
DOIs
StatePublished - 2008
Event2008 IEEE International Conference on Image Processing, ICIP 2008 - San Diego, CA, United States
Duration: Oct 12 2008Oct 15 2008

Other

Other2008 IEEE International Conference on Image Processing, ICIP 2008
CountryUnited States
CitySan Diego, CA
Period10/12/0810/15/08

Fingerprint

Image denoising
Covariance matrix

Keywords

  • Gaussian scale mixture
  • Image denoising
  • Image modelling

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Guerrero-Colón, J. A., Simoncelli, E., & Portilla, J. (2008). Image denoising using mixtures of gaussian scale mixtures. In 2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings (pp. 565-568). [4711817] https://doi.org/10.1109/ICIP.2008.4711817

Image denoising using mixtures of gaussian scale mixtures. / Guerrero-Colón, Jose A.; Simoncelli, Eero; Portilla, Javier.

2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings. 2008. p. 565-568 4711817.

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

Guerrero-Colón, JA, Simoncelli, E & Portilla, J 2008, Image denoising using mixtures of gaussian scale mixtures. in 2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings., 4711817, pp. 565-568, 2008 IEEE International Conference on Image Processing, ICIP 2008, San Diego, CA, United States, 10/12/08. https://doi.org/10.1109/ICIP.2008.4711817
Guerrero-Colón JA, Simoncelli E, Portilla J. Image denoising using mixtures of gaussian scale mixtures. In 2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings. 2008. p. 565-568. 4711817 https://doi.org/10.1109/ICIP.2008.4711817
Guerrero-Colón, Jose A. ; Simoncelli, Eero ; Portilla, Javier. / Image denoising using mixtures of gaussian scale mixtures. 2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings. 2008. pp. 565-568
@inproceedings{6daa1361249147fe935ab39ffc53aea7,
title = "Image denoising using mixtures of gaussian scale mixtures",
abstract = "The local statistical properties of photographic images, when represented in a multi-scale basis, have been described using Gaussian scale mixtures (GSMs). In that model, each spatial neighborhood of coefficients is described as a Gaussian random vector modulated by a random hidden positive scaling variable. Here, we introduce a more powerful model in which neighborhoods of each subband are described as a finite mixture of GSMs. We develop methods to learn the mixing densities and covariance matrices associated with each of the GSM components from a single image, and show that this process naturally segments the image into regions of similar content. The model parameters can also be learned in the presence of additive Gaussian noise, and the resulting fitted model may be used as a prior for Bayesian noise removal. Simulations demonstrate this model substantially outperforms the original GSM model.",
keywords = "Gaussian scale mixture, Image denoising, Image modelling",
author = "Guerrero-Col{\'o}n, {Jose A.} and Eero Simoncelli and Javier Portilla",
year = "2008",
doi = "10.1109/ICIP.2008.4711817",
language = "English (US)",
isbn = "1424417643",
pages = "565--568",
booktitle = "2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings",

}

TY - GEN

T1 - Image denoising using mixtures of gaussian scale mixtures

AU - Guerrero-Colón, Jose A.

AU - Simoncelli, Eero

AU - Portilla, Javier

PY - 2008

Y1 - 2008

N2 - The local statistical properties of photographic images, when represented in a multi-scale basis, have been described using Gaussian scale mixtures (GSMs). In that model, each spatial neighborhood of coefficients is described as a Gaussian random vector modulated by a random hidden positive scaling variable. Here, we introduce a more powerful model in which neighborhoods of each subband are described as a finite mixture of GSMs. We develop methods to learn the mixing densities and covariance matrices associated with each of the GSM components from a single image, and show that this process naturally segments the image into regions of similar content. The model parameters can also be learned in the presence of additive Gaussian noise, and the resulting fitted model may be used as a prior for Bayesian noise removal. Simulations demonstrate this model substantially outperforms the original GSM model.

AB - The local statistical properties of photographic images, when represented in a multi-scale basis, have been described using Gaussian scale mixtures (GSMs). In that model, each spatial neighborhood of coefficients is described as a Gaussian random vector modulated by a random hidden positive scaling variable. Here, we introduce a more powerful model in which neighborhoods of each subband are described as a finite mixture of GSMs. We develop methods to learn the mixing densities and covariance matrices associated with each of the GSM components from a single image, and show that this process naturally segments the image into regions of similar content. The model parameters can also be learned in the presence of additive Gaussian noise, and the resulting fitted model may be used as a prior for Bayesian noise removal. Simulations demonstrate this model substantially outperforms the original GSM model.

KW - Gaussian scale mixture

KW - Image denoising

KW - Image modelling

UR - http://www.scopus.com/inward/record.url?scp=68249125631&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=68249125631&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2008.4711817

DO - 10.1109/ICIP.2008.4711817

M3 - Conference contribution

AN - SCOPUS:68249125631

SN - 1424417643

SN - 9781424417643

SP - 565

EP - 568

BT - 2008 IEEE International Conference on Image Processing, ICIP 2008 Proceedings

ER -