Item type |
SIG Technical Reports(1) |
公開日 |
2020-02-20 |
タイトル |
|
|
タイトル |
A Study on Degraded Image Super-Resolution with Joint De-artifact and Deblurring |
タイトル |
|
|
言語 |
en |
|
タイトル |
A Study on Degraded Image Super-Resolution with Joint De-artifact and Deblurring |
言語 |
|
|
言語 |
eng |
資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_18gh |
|
資源タイプ |
technical report |
著者所属 |
|
|
|
KDDI Research, Inc. |
著者所属 |
|
|
|
KDDI Research, Inc. |
著者所属(英) |
|
|
|
en |
|
|
KDDI Research, Inc. |
著者所属(英) |
|
|
|
en |
|
|
KDDI Research, Inc. |
著者名 |
Jianfeng, Xu
Kazuyuki, Tasaka
|
著者名(英) |
Jianfeng, Xu
Kazuyuki, Tasaka
|
論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
In this paper, we address the degraded image super-resolution problem in a multi-task learning (MTL) manner. An all-in-one collaboration framework (ACF) with a learnable junction unit has been proposed to handle two major problems that exist in MTL: How to share and How much to share. Specifically, considering the intrinsic characteristic of multiple image degradations, we propose to first deal with the compression artifact, motion blur (dynamic blur), and spatial structure information of the input image in parallel under a three-branch architecture. Then, based on the feedback from the final super-resolution reconstruction, we train a learnable junction unit with a dual-voting mechanism to selectively filter or preserve these shared feature representations. This unit facilitates collaboration between multi-tasks and learns an optimal combination of different features. Experimental results show that the proposed all-in-one collaboration framework not only produce favorable high-resolution results while removing serious degradation, but also has high computational efficiency, outperforming state-of-the-art methods. |
論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
In this paper, we address the degraded image super-resolution problem in a multi-task learning (MTL) manner. An all-in-one collaboration framework (ACF) with a learnable junction unit has been proposed to handle two major problems that exist in MTL: How to share and How much to share. Specifically, considering the intrinsic characteristic of multiple image degradations, we propose to first deal with the compression artifact, motion blur (dynamic blur), and spatial structure information of the input image in parallel under a three-branch architecture. Then, based on the feedback from the final super-resolution reconstruction, we train a learnable junction unit with a dual-voting mechanism to selectively filter or preserve these shared feature representations. This unit facilitates collaboration between multi-tasks and learns an optimal combination of different features. Experimental results show that the proposed all-in-one collaboration framework not only produce favorable high-resolution results while removing serious degradation, but also has high computational efficiency, outperforming state-of-the-art methods. |
書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AN10438399 |
書誌情報 |
研究報告オーディオビジュアル複合情報処理(AVM)
巻 2020-AVM-108,
号 11,
p. 1-6,
発行日 2020-02-20
|
ISSN |
|
|
収録物識別子タイプ |
ISSN |
|
収録物識別子 |
2188-8582 |
Notice |
|
|
|
SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc. |
出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |