Abstract

Image super-resolution (SR) is a challenging task which aims to recover the high-resolution (HR) images from the degraded low-resolution (LR) observations. To address this ill-posed problem, properly exploiting the image prior is of great importance. In this paper, we propose a data-adaptive low-rank (DLR) model. Rather than directly assuming that the rank of a group of similar patches is low, the DLR model imposes the low-rank property on the residual of the grouped patches. In addition, the shape of the patches in our DLR model is adapted to the contents of images, so that the dissimilar pixels in a group of patches can be largely reduced. In order to further boost the performance, an external gradient prior (EGP), which is learned externally to capture gradient information, is combined with DLR to form a joint prior. When solving the DLR-based and the joint-prior-based minimization problems, the split Bregman method is adopted to speed up the convergence. The extensive experimental results show that our algorithms outperform many state-of-the-art single image SR methods in terms of both objective and subjective qualities.

Original languageEnglish (US)
Pages (from-to)36-49
Number of pages14
JournalSignal Processing
Volume161
DOIs
StatePublished - Aug 1 2019

Fingerprint

Optical resolving power
Image resolution
Pixels

Keywords

  • Gradient prior
  • Low-rank modeling
  • Split Bregman method
  • Steering kernel
  • Super-resolution

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Electrical and Electronic Engineering

Cite this

Data-adaptive low-rank modeling and external gradient prior for single image super-resolution. / Chang, Kan; Zhang, Xueyu; Ding, Pak Lun Kevin; Li, Baoxin.

In: Signal Processing, Vol. 161, 01.08.2019, p. 36-49.

Research output: Contribution to journalArticle

Chang, Kan ; Zhang, Xueyu ; Ding, Pak Lun Kevin ; Li, Baoxin. / Data-adaptive low-rank modeling and external gradient prior for single image super-resolution. In: Signal Processing. 2019 ; Vol. 161. pp. 36-49.
@article{091663ff18c845f981ce1a44d2b641e7,
title = "Data-adaptive low-rank modeling and external gradient prior for single image super-resolution",
abstract = "Image super-resolution (SR) is a challenging task which aims to recover the high-resolution (HR) images from the degraded low-resolution (LR) observations. To address this ill-posed problem, properly exploiting the image prior is of great importance. In this paper, we propose a data-adaptive low-rank (DLR) model. Rather than directly assuming that the rank of a group of similar patches is low, the DLR model imposes the low-rank property on the residual of the grouped patches. In addition, the shape of the patches in our DLR model is adapted to the contents of images, so that the dissimilar pixels in a group of patches can be largely reduced. In order to further boost the performance, an external gradient prior (EGP), which is learned externally to capture gradient information, is combined with DLR to form a joint prior. When solving the DLR-based and the joint-prior-based minimization problems, the split Bregman method is adopted to speed up the convergence. The extensive experimental results show that our algorithms outperform many state-of-the-art single image SR methods in terms of both objective and subjective qualities.",
keywords = "Gradient prior, Low-rank modeling, Split Bregman method, Steering kernel, Super-resolution",
author = "Kan Chang and Xueyu Zhang and Ding, {Pak Lun Kevin} and Baoxin Li",
year = "2019",
month = "8",
day = "1",
doi = "10.1016/j.sigpro.2019.03.011",
language = "English (US)",
volume = "161",
pages = "36--49",
journal = "Signal Processing",
issn = "0165-1684",
publisher = "Elsevier",

}

TY - JOUR

T1 - Data-adaptive low-rank modeling and external gradient prior for single image super-resolution

AU - Chang, Kan

AU - Zhang, Xueyu

AU - Ding, Pak Lun Kevin

AU - Li, Baoxin

PY - 2019/8/1

Y1 - 2019/8/1

N2 - Image super-resolution (SR) is a challenging task which aims to recover the high-resolution (HR) images from the degraded low-resolution (LR) observations. To address this ill-posed problem, properly exploiting the image prior is of great importance. In this paper, we propose a data-adaptive low-rank (DLR) model. Rather than directly assuming that the rank of a group of similar patches is low, the DLR model imposes the low-rank property on the residual of the grouped patches. In addition, the shape of the patches in our DLR model is adapted to the contents of images, so that the dissimilar pixels in a group of patches can be largely reduced. In order to further boost the performance, an external gradient prior (EGP), which is learned externally to capture gradient information, is combined with DLR to form a joint prior. When solving the DLR-based and the joint-prior-based minimization problems, the split Bregman method is adopted to speed up the convergence. The extensive experimental results show that our algorithms outperform many state-of-the-art single image SR methods in terms of both objective and subjective qualities.

AB - Image super-resolution (SR) is a challenging task which aims to recover the high-resolution (HR) images from the degraded low-resolution (LR) observations. To address this ill-posed problem, properly exploiting the image prior is of great importance. In this paper, we propose a data-adaptive low-rank (DLR) model. Rather than directly assuming that the rank of a group of similar patches is low, the DLR model imposes the low-rank property on the residual of the grouped patches. In addition, the shape of the patches in our DLR model is adapted to the contents of images, so that the dissimilar pixels in a group of patches can be largely reduced. In order to further boost the performance, an external gradient prior (EGP), which is learned externally to capture gradient information, is combined with DLR to form a joint prior. When solving the DLR-based and the joint-prior-based minimization problems, the split Bregman method is adopted to speed up the convergence. The extensive experimental results show that our algorithms outperform many state-of-the-art single image SR methods in terms of both objective and subjective qualities.

KW - Gradient prior

KW - Low-rank modeling

KW - Split Bregman method

KW - Steering kernel

KW - Super-resolution

UR - http://www.scopus.com/inward/record.url?scp=85063049623&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063049623&partnerID=8YFLogxK

U2 - 10.1016/j.sigpro.2019.03.011

DO - 10.1016/j.sigpro.2019.03.011

M3 - Article

AN - SCOPUS:85063049623

VL - 161

SP - 36

EP - 49

JO - Signal Processing

JF - Signal Processing

SN - 0165-1684

ER -