Volume 13 Number 11 (Nov. 2018)
Home > Archive > 2018 > Volume 13 Number 11 (Nov. 2018) >
JCP 2018 Vol.13(11): 1246-1264 ISSN: 1796-203X
doi: 10.17706/jcp.13.11.1246-1264

Reverse Sparse Representation for Single Image Super-Resolution

Jian Xu1, 2, 3, Chunyu Wang1, 2, 3, Jiulun Fan1, 2, 3
1Key Laboratory of Electronic Information Application Technology for Scene Investigation, Ministry of Public Security, Xi'an, 710121, China.
2International Joint Research Center for Wireless Communication and Information Processing, Shaanxi, Xi'an, 710121, China.
3School of Telecommunication and Information Engineering, Xi’an University of Posts and Telecommunications, Xi’an 710121, China.

Abstract—The key issue facing the machine learning based super-resolution (SR) method is how to describe the relationship between the low-resolution (LR) and high-resolution (HR) images. Sparse representation techniques have provided effective tools for this task. In classical coupled dictionary models, the most important issue is how to train two dictionaries to convert the HR and LR data samples to a unified feature subspace. To address this problem, this paper presents novel coupled dictionary training approach for SR. In the proposed model, reverse sparse representation constrains are employed to train coupled dictionaries to reduce the weaknesses of the SR problem. To avoid the alternative iteration and reduce the time complexity, the HR and LR dictionaries are trained in two steps. First, the HR dictionary is trained with the traditional single dictionary training algorithm. Next, according to the HR dictionary and the HR data set, the reverse sparse representations are prepared to generate the LR atoms. Finally, the LR dictionary is generated with reverse sparse representations and the LR data set. Experimental results demonstrate that our approach outperforms 7 related approaches.

Index Terms—Super-resolution, sparse representation, dictionary training, non local mean regularization.


Cite: Jian Xu, Chunyu Wang, Jiulun Fan, "Reverse Sparse Representation for Single Image Super-Resolution," Journal of Computers vol. 13, no. 11, pp. 1246-1264 , 2018.

General Information

ISSN: 1796-203X
Abbreviated Title: J.Comput.
Frequency: Bimonthly
Editor-in-Chief: Prof. Liansheng Tan
Executive Editor: Ms. Nina Lee
Abstracting/ Indexing: DBLP, EBSCO,  ProQuest, INSPEC, ULRICH's Periodicals Directory, WorldCat,etc
E-mail: jcp@iap.org
  • Nov 14, 2019 News!

    Vol 14, No 11 has been published with online version   [Click]

  • Mar 20, 2020 News!

    Vol 15, No 2 has been published with online version   [Click]

  • Dec 16, 2019 News!

    Vol 14, No 12 has been published with online version   [Click]

  • Sep 16, 2019 News!

    Vol 14, No 9 has been published with online version   [Click]

  • Aug 16, 2019 News!

    Vol 14, No 8 has been published with online version   [Click]

  • Read more>>