Abstract

This letter proposes a reconstruction-based single image super resolution (SR) method by using joint regularization, where a group-residual-based regularization (GRR) and a ridge-regression-based regularization (3R) are combined. In GRR, non-local similar patches are grouped together, and the group weights are calculated so as to adaptively constrain the residual values in the gradient domain. In 3R, we adopt the ridge-regression-based method to establish the projection matrices from an external high-resolution (HR) training set, so that the external HR information can be utilized. To obtain an estimation of the targeted HR image, an efficient algorithm is designed for solving the joint formulation. Experimental results on different image datasets indicate that the proposed method is able to achieve the state-of-the-art performance.

Original languageEnglish (US)
JournalIEEE Signal Processing Letters
DOIs
StateAccepted/In press - Mar 10 2018

Fingerprint

Super-resolution
Image resolution
Regularization
Ridge Regression
High Resolution
Projection Matrix
Patch
Efficient Algorithms
Gradient
Formulation
Experimental Results

Keywords

  • Non-local Self-similarity
  • Regularization
  • Ridge Regression
  • Super Resolution
  • Total Variation

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering
  • Applied Mathematics

Cite this

Single Image Super Resolution Using Joint Regularization. / Chang, Kan; Ding, Pak Lun Kevin; Li, Baoxin.

In: IEEE Signal Processing Letters, 10.03.2018.

Research output: Contribution to journalArticle

@article{fe5e8c65703f4bb3b1e41d5512f7a873,
title = "Single Image Super Resolution Using Joint Regularization",
abstract = "This letter proposes a reconstruction-based single image super resolution (SR) method by using joint regularization, where a group-residual-based regularization (GRR) and a ridge-regression-based regularization (3R) are combined. In GRR, non-local similar patches are grouped together, and the group weights are calculated so as to adaptively constrain the residual values in the gradient domain. In 3R, we adopt the ridge-regression-based method to establish the projection matrices from an external high-resolution (HR) training set, so that the external HR information can be utilized. To obtain an estimation of the targeted HR image, an efficient algorithm is designed for solving the joint formulation. Experimental results on different image datasets indicate that the proposed method is able to achieve the state-of-the-art performance.",
keywords = "Non-local Self-similarity, Regularization, Ridge Regression, Super Resolution, Total Variation",
author = "Kan Chang and Ding, {Pak Lun Kevin} and Baoxin Li",
year = "2018",
month = "3",
day = "10",
doi = "10.1109/LSP.2018.2815003",
language = "English (US)",
journal = "IEEE Signal Processing Letters",
issn = "1070-9908",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Single Image Super Resolution Using Joint Regularization

AU - Chang, Kan

AU - Ding, Pak Lun Kevin

AU - Li, Baoxin

PY - 2018/3/10

Y1 - 2018/3/10

N2 - This letter proposes a reconstruction-based single image super resolution (SR) method by using joint regularization, where a group-residual-based regularization (GRR) and a ridge-regression-based regularization (3R) are combined. In GRR, non-local similar patches are grouped together, and the group weights are calculated so as to adaptively constrain the residual values in the gradient domain. In 3R, we adopt the ridge-regression-based method to establish the projection matrices from an external high-resolution (HR) training set, so that the external HR information can be utilized. To obtain an estimation of the targeted HR image, an efficient algorithm is designed for solving the joint formulation. Experimental results on different image datasets indicate that the proposed method is able to achieve the state-of-the-art performance.

AB - This letter proposes a reconstruction-based single image super resolution (SR) method by using joint regularization, where a group-residual-based regularization (GRR) and a ridge-regression-based regularization (3R) are combined. In GRR, non-local similar patches are grouped together, and the group weights are calculated so as to adaptively constrain the residual values in the gradient domain. In 3R, we adopt the ridge-regression-based method to establish the projection matrices from an external high-resolution (HR) training set, so that the external HR information can be utilized. To obtain an estimation of the targeted HR image, an efficient algorithm is designed for solving the joint formulation. Experimental results on different image datasets indicate that the proposed method is able to achieve the state-of-the-art performance.

KW - Non-local Self-similarity

KW - Regularization

KW - Ridge Regression

KW - Super Resolution

KW - Total Variation

UR - http://www.scopus.com/inward/record.url?scp=85043466476&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85043466476&partnerID=8YFLogxK

U2 - 10.1109/LSP.2018.2815003

DO - 10.1109/LSP.2018.2815003

M3 - Article

JO - IEEE Signal Processing Letters

JF - IEEE Signal Processing Letters

SN - 1070-9908

ER -