Low-rank total variation for image super-resolution

Feng Shi, Jian Cheng, Li Wang, Pew Thian Yap, Dinggang Shen

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

19 Citations (Scopus)

Abstract

Most natural images can be approximated using their low-rank components. This fact has been successfully exploited in recent advancements of matrix completion algorithms for image recovery. However, a major limitation of low-rank matrix completion algorithms is that they cannot recover the case where a whole row or column is missing. The missing row or column will be simply filled as an arbitrary combination of other rows or columns with known values. This precludes the application of matrix completion to problems such as super-resolution (SR) where missing values in many rows and columns need to be recovered in the process of up-sampling a low-resolution image. Moreover, low-rank regularization considers information globally from the whole image and does not take proper consideration of local spatial consistency. Accordingly, we propose in this paper a solution to the SR problem via simultaneous (global) low-rank and (local) total variation (TV) regularization. We solve the respective cost function using the alternating direction method of multipliers (ADMM). Experiments on MR images of adults and pediatric subjects demonstrate that the proposed method enhances the details of the recovered high-resolution images, and outperforms the nearest-neighbor interpolation, cubic interpolation, non-local means, and TV-based up-sampling.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages155-162
Number of pages8
Volume8149 LNCS
EditionPART 1
DOIs
Publication statusPublished - 2013 Oct 23
Externally publishedYes
Event16th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2013 - Nagoya, Japan
Duration: 2013 Sep 222013 Sep 26

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume8149 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other16th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2013
CountryJapan
CityNagoya
Period13/9/2213/9/26

Fingerprint

Super-resolution
Total Variation
Matrix Completion
Image resolution
Interpolation
Sampling
Pediatrics
Cost functions
Interpolate
Image Recovery
Total Variation Regularization
Method of multipliers
Low-rank Matrices
Alternating Direction Method
Missing Values
Recovery
Cost Function
Nearest Neighbor
Regularization
High Resolution

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Shi, F., Cheng, J., Wang, L., Yap, P. T., & Shen, D. (2013). Low-rank total variation for image super-resolution. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 1 ed., Vol. 8149 LNCS, pp. 155-162). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8149 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-40811-3_20

Low-rank total variation for image super-resolution. / Shi, Feng; Cheng, Jian; Wang, Li; Yap, Pew Thian; Shen, Dinggang.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8149 LNCS PART 1. ed. 2013. p. 155-162 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8149 LNCS, No. PART 1).

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

Shi, F, Cheng, J, Wang, L, Yap, PT & Shen, D 2013, Low-rank total variation for image super-resolution. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 edn, vol. 8149 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 8149 LNCS, pp. 155-162, 16th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2013, Nagoya, Japan, 13/9/22. https://doi.org/10.1007/978-3-642-40811-3_20
Shi F, Cheng J, Wang L, Yap PT, Shen D. Low-rank total variation for image super-resolution. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 ed. Vol. 8149 LNCS. 2013. p. 155-162. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-642-40811-3_20
Shi, Feng ; Cheng, Jian ; Wang, Li ; Yap, Pew Thian ; Shen, Dinggang. / Low-rank total variation for image super-resolution. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8149 LNCS PART 1. ed. 2013. pp. 155-162 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).
@inproceedings{e73d7348f805474bba99be2adc4b14cf,
title = "Low-rank total variation for image super-resolution",
abstract = "Most natural images can be approximated using their low-rank components. This fact has been successfully exploited in recent advancements of matrix completion algorithms for image recovery. However, a major limitation of low-rank matrix completion algorithms is that they cannot recover the case where a whole row or column is missing. The missing row or column will be simply filled as an arbitrary combination of other rows or columns with known values. This precludes the application of matrix completion to problems such as super-resolution (SR) where missing values in many rows and columns need to be recovered in the process of up-sampling a low-resolution image. Moreover, low-rank regularization considers information globally from the whole image and does not take proper consideration of local spatial consistency. Accordingly, we propose in this paper a solution to the SR problem via simultaneous (global) low-rank and (local) total variation (TV) regularization. We solve the respective cost function using the alternating direction method of multipliers (ADMM). Experiments on MR images of adults and pediatric subjects demonstrate that the proposed method enhances the details of the recovered high-resolution images, and outperforms the nearest-neighbor interpolation, cubic interpolation, non-local means, and TV-based up-sampling.",
author = "Feng Shi and Jian Cheng and Li Wang and Yap, {Pew Thian} and Dinggang Shen",
year = "2013",
month = "10",
day = "23",
doi = "10.1007/978-3-642-40811-3_20",
language = "English",
isbn = "9783642408106",
volume = "8149 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 1",
pages = "155--162",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 1",

}

TY - GEN

T1 - Low-rank total variation for image super-resolution

AU - Shi, Feng

AU - Cheng, Jian

AU - Wang, Li

AU - Yap, Pew Thian

AU - Shen, Dinggang

PY - 2013/10/23

Y1 - 2013/10/23

N2 - Most natural images can be approximated using their low-rank components. This fact has been successfully exploited in recent advancements of matrix completion algorithms for image recovery. However, a major limitation of low-rank matrix completion algorithms is that they cannot recover the case where a whole row or column is missing. The missing row or column will be simply filled as an arbitrary combination of other rows or columns with known values. This precludes the application of matrix completion to problems such as super-resolution (SR) where missing values in many rows and columns need to be recovered in the process of up-sampling a low-resolution image. Moreover, low-rank regularization considers information globally from the whole image and does not take proper consideration of local spatial consistency. Accordingly, we propose in this paper a solution to the SR problem via simultaneous (global) low-rank and (local) total variation (TV) regularization. We solve the respective cost function using the alternating direction method of multipliers (ADMM). Experiments on MR images of adults and pediatric subjects demonstrate that the proposed method enhances the details of the recovered high-resolution images, and outperforms the nearest-neighbor interpolation, cubic interpolation, non-local means, and TV-based up-sampling.

AB - Most natural images can be approximated using their low-rank components. This fact has been successfully exploited in recent advancements of matrix completion algorithms for image recovery. However, a major limitation of low-rank matrix completion algorithms is that they cannot recover the case where a whole row or column is missing. The missing row or column will be simply filled as an arbitrary combination of other rows or columns with known values. This precludes the application of matrix completion to problems such as super-resolution (SR) where missing values in many rows and columns need to be recovered in the process of up-sampling a low-resolution image. Moreover, low-rank regularization considers information globally from the whole image and does not take proper consideration of local spatial consistency. Accordingly, we propose in this paper a solution to the SR problem via simultaneous (global) low-rank and (local) total variation (TV) regularization. We solve the respective cost function using the alternating direction method of multipliers (ADMM). Experiments on MR images of adults and pediatric subjects demonstrate that the proposed method enhances the details of the recovered high-resolution images, and outperforms the nearest-neighbor interpolation, cubic interpolation, non-local means, and TV-based up-sampling.

UR - http://www.scopus.com/inward/record.url?scp=84885770677&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84885770677&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-40811-3_20

DO - 10.1007/978-3-642-40811-3_20

M3 - Conference contribution

C2 - 24505661

AN - SCOPUS:84885770677

SN - 9783642408106

VL - 8149 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 155

EP - 162

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -