Abstract
The recent years have witnessed significant advances in image deblurring. In general, the success of deblurring methods depends heavily on extraction of salient structures from a blurry image for kernel estimation. Most deblurring methods often operate on high-resolution images where contours or edges can be extracted for kernel estimation. However, recovering reliable structures from low-resolution images becomes extremely challenging. In this paper, we propose a spatially variant deblurring algorithm for low-resolution images based on the exemplars. To exploit the exemplar information, we develop a super-resolution guided method to help the restoration of reliable image structures which can be used for kernel estimation. Experimental evaluations against the state-of-the-art methods show that the proposed algorithm performs favorably in deblurring low-resolution images. Furthermore, we show that the SR results obtained as byproducts in our method are comparable compared to other blind SR methods.
Original language | English |
---|---|
Title of host publication | Computer Vision - ACCV 2016 Workshops - ACCV 2016 International Workshops, Revised Selected Papers |
Editors | Jiwen Lu, Kai-Kuang Ma, Chu-Song Chen |
Publisher | Springer Verlag |
Pages | 111-127 |
Number of pages | 17 |
ISBN (Print) | 9783319544069 |
DOIs | |
Publication status | Published - 2017 |
Event | 13th Asian Conference on Computer Vision, ACCV 2016 - Taipei, Taiwan, Province of China Duration: 2016 Nov 20 → 2016 Nov 24 |
Publication series
Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|
Volume | 10116 LNCS |
ISSN (Print) | 0302-9743 |
ISSN (Electronic) | 1611-3349 |
Other
Other | 13th Asian Conference on Computer Vision, ACCV 2016 |
---|---|
Country | Taiwan, Province of China |
City | Taipei |
Period | 16/11/20 → 16/11/24 |
Bibliographical note
Funding Information:This work has been supported in part by NSF CAREER (No. 1149783), NSF IIS (No. 1152576), NSFC (No. 61572099 and 61320106008) and a gift from Adobe.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)