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.
|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|
|Number of pages||17|
|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
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Other||13th Asian Conference on Computer Vision, ACCV 2016|
|Country||Taiwan, Province of China|
|Period||16/11/20 → 16/11/24|
Bibliographical noteFunding 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)