Example-driven manifold priors for image deconvolution

Jie Ni, Pavan Turaga, Vishal Patel, Rama Chellappa

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

Image restoration methods that exploit prior information about images to be estimated have been extensively studied, typically using the Bayesian framework. In this paper, we consider the role of prior knowledge of the object class in the form of a patch manifold to address the deconvolution problem. Specifically, we incorporate unlabeled image data of the object class, say natural images, in the form of a patch-manifold prior for the object class. The manifold prior is implicitly estimated from the given unlabeled data. We show how the patch-manifold prior effectively exploits the available sample class data for regularizing the deblurring problem. Furthermore, we derive a generalized cross-validation (GCV) function to automatically determine the regularization parameter at each iteration without explicitly knowing the noise variance. Extensive experiments show that this method performs better than many competitive image deconvolution methods.

Original languageEnglish (US)
Article number5753939
Pages (from-to)3086-3096
Number of pages11
JournalIEEE Transactions on Image Processing
Volume20
Issue number11
DOIs
StatePublished - Nov 1 2011

Fingerprint

Deconvolution
Image reconstruction
Experiments

All Science Journal Classification (ASJC) codes

  • Software
  • Computer Graphics and Computer-Aided Design

Cite this

Ni, J., Turaga, P., Patel, V., & Chellappa, R. (2011). Example-driven manifold priors for image deconvolution. IEEE Transactions on Image Processing, 20(11), 3086-3096. [5753939]. https://doi.org/10.1109/TIP.2011.2145386
Ni, Jie ; Turaga, Pavan ; Patel, Vishal ; Chellappa, Rama. / Example-driven manifold priors for image deconvolution. In: IEEE Transactions on Image Processing. 2011 ; Vol. 20, No. 11. pp. 3086-3096.
@article{69bf9ad649bc46dfba1217bf1b309256,
title = "Example-driven manifold priors for image deconvolution",
abstract = "Image restoration methods that exploit prior information about images to be estimated have been extensively studied, typically using the Bayesian framework. In this paper, we consider the role of prior knowledge of the object class in the form of a patch manifold to address the deconvolution problem. Specifically, we incorporate unlabeled image data of the object class, say natural images, in the form of a patch-manifold prior for the object class. The manifold prior is implicitly estimated from the given unlabeled data. We show how the patch-manifold prior effectively exploits the available sample class data for regularizing the deblurring problem. Furthermore, we derive a generalized cross-validation (GCV) function to automatically determine the regularization parameter at each iteration without explicitly knowing the noise variance. Extensive experiments show that this method performs better than many competitive image deconvolution methods.",
author = "Jie Ni and Pavan Turaga and Vishal Patel and Rama Chellappa",
year = "2011",
month = "11",
day = "1",
doi = "10.1109/TIP.2011.2145386",
language = "English (US)",
volume = "20",
pages = "3086--3096",
journal = "IEEE Transactions on Image Processing",
issn = "1057-7149",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "11",

}

Ni, J, Turaga, P, Patel, V & Chellappa, R 2011, 'Example-driven manifold priors for image deconvolution', IEEE Transactions on Image Processing, vol. 20, no. 11, 5753939, pp. 3086-3096. https://doi.org/10.1109/TIP.2011.2145386

Example-driven manifold priors for image deconvolution. / Ni, Jie; Turaga, Pavan; Patel, Vishal; Chellappa, Rama.

In: IEEE Transactions on Image Processing, Vol. 20, No. 11, 5753939, 01.11.2011, p. 3086-3096.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Example-driven manifold priors for image deconvolution

AU - Ni, Jie

AU - Turaga, Pavan

AU - Patel, Vishal

AU - Chellappa, Rama

PY - 2011/11/1

Y1 - 2011/11/1

N2 - Image restoration methods that exploit prior information about images to be estimated have been extensively studied, typically using the Bayesian framework. In this paper, we consider the role of prior knowledge of the object class in the form of a patch manifold to address the deconvolution problem. Specifically, we incorporate unlabeled image data of the object class, say natural images, in the form of a patch-manifold prior for the object class. The manifold prior is implicitly estimated from the given unlabeled data. We show how the patch-manifold prior effectively exploits the available sample class data for regularizing the deblurring problem. Furthermore, we derive a generalized cross-validation (GCV) function to automatically determine the regularization parameter at each iteration without explicitly knowing the noise variance. Extensive experiments show that this method performs better than many competitive image deconvolution methods.

AB - Image restoration methods that exploit prior information about images to be estimated have been extensively studied, typically using the Bayesian framework. In this paper, we consider the role of prior knowledge of the object class in the form of a patch manifold to address the deconvolution problem. Specifically, we incorporate unlabeled image data of the object class, say natural images, in the form of a patch-manifold prior for the object class. The manifold prior is implicitly estimated from the given unlabeled data. We show how the patch-manifold prior effectively exploits the available sample class data for regularizing the deblurring problem. Furthermore, we derive a generalized cross-validation (GCV) function to automatically determine the regularization parameter at each iteration without explicitly knowing the noise variance. Extensive experiments show that this method performs better than many competitive image deconvolution methods.

UR - http://www.scopus.com/inward/record.url?scp=80054805330&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80054805330&partnerID=8YFLogxK

U2 - 10.1109/TIP.2011.2145386

DO - 10.1109/TIP.2011.2145386

M3 - Article

C2 - 21518663

AN - SCOPUS:80054805330

VL - 20

SP - 3086

EP - 3096

JO - IEEE Transactions on Image Processing

JF - IEEE Transactions on Image Processing

SN - 1057-7149

IS - 11

M1 - 5753939

ER -