Training compression artifacts reduction network with domain adaptation

Yu Jin Ham; Chaehwa Yoo; Je Won Kang

doi:10.1117/12.2597094

Training compression artifacts reduction network with domain adaptation

Yu Jin Ham, Chaehwa Yoo, Je Won Kang

Electronic and Electrical Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Scopus citations

Abstract

Compression artifact removal is imperative for more visually pleasing contents after image and video compression. Recent works on compression artifact reduction network (CARN) assume that the same or similar quality of images would be employed for both training and testing, and, accordingly, a model needs a quality factor as a prior to accomplish the task successfully. However, the possible discrepancy will degrade performance substantially in a target if the model confronts a different level of distortion from the training phase. To solve the problem, we propose a novel training scheme of CARN to take an advantage of domain adaptation (DA). Specifically, we assign an image encoded with a different quality factor as a different domain and train a CARN using DA to perform robustly in another domain of a different level of distortion. Experimental results demonstrate that the proposed method achieves superior performance on DIV2K, BSD68, and Set12.

Original language	English
Title of host publication	Applications of Digital Image Processing XLIV
Editors	Andrew G. Tescher, Touradj Ebrahimi
Publisher	SPIE
ISBN (Electronic)	9781510645226
DOIs	https://doi.org/10.1117/12.2597094
State	Published - 2021
Event	Applications of Digital Image Processing XLIV 2021 - San Diego, United States Duration: 1 Aug 2021 → 5 Aug 2021

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	11842
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	Applications of Digital Image Processing XLIV 2021
Country/Territory	United States
City	San Diego
Period	1/08/21 → 5/08/21

Bibliographical note

Funding Information:
This research was supported by the MSIT(Ministry of Science and ICT), Korea, under the ITRC(Information Technology Research Center) support program(IITP–2021–2020–0–01460) supervised by the IITP(Institute for Information & Communications Technology Planning & Evaluation)

Publisher Copyright:
© COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.

Keywords

Compression Artifact Reduction
Deep Learning
Domain Adaptation
Video/Image Coding

Access to Document

10.1117/12.2597094

Cite this

@inproceedings{d5ac07973eed424c858117ba6764a148,

title = "Training compression artifacts reduction network with domain adaptation",

abstract = "Compression artifact removal is imperative for more visually pleasing contents after image and video compression. Recent works on compression artifact reduction network (CARN) assume that the same or similar quality of images would be employed for both training and testing, and, accordingly, a model needs a quality factor as a prior to accomplish the task successfully. However, the possible discrepancy will degrade performance substantially in a target if the model confronts a different level of distortion from the training phase. To solve the problem, we propose a novel training scheme of CARN to take an advantage of domain adaptation (DA). Specifically, we assign an image encoded with a different quality factor as a different domain and train a CARN using DA to perform robustly in another domain of a different level of distortion. Experimental results demonstrate that the proposed method achieves superior performance on DIV2K, BSD68, and Set12. ",

keywords = "Compression Artifact Reduction, Deep Learning, Domain Adaptation, Video/Image Coding",

author = "Ham, {Yu Jin} and Chaehwa Yoo and Kang, {Je Won}",

note = "Funding Information: This research was supported by the MSIT(Ministry of Science and ICT), Korea, under the ITRC(Information Technology Research Center) support program(IITP–2021–2020–0–01460) supervised by the IITP(Institute for Information & Communications Technology Planning & Evaluation) Publisher Copyright: {\textcopyright} COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.; Applications of Digital Image Processing XLIV 2021 ; Conference date: 01-08-2021 Through 05-08-2021",

year = "2021",

doi = "10.1117/12.2597094",

language = "English",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

publisher = "SPIE",

editor = "Tescher, {Andrew G.} and Touradj Ebrahimi",

booktitle = "Applications of Digital Image Processing XLIV",

}

Ham, YJ, Yoo, C & Kang, JW 2021, Training compression artifacts reduction network with domain adaptation. in AG Tescher & T Ebrahimi (eds), Applications of Digital Image Processing XLIV., 118420U, Proceedings of SPIE - The International Society for Optical Engineering, vol. 11842, SPIE, Applications of Digital Image Processing XLIV 2021, San Diego, United States, 1/08/21. https://doi.org/10.1117/12.2597094

Training compression artifacts reduction network with domain adaptation. / Ham, Yu Jin; Yoo, Chaehwa; Kang, Je Won.
Applications of Digital Image Processing XLIV. ed. / Andrew G. Tescher; Touradj Ebrahimi. SPIE, 2021. 118420U (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 11842).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Training compression artifacts reduction network with domain adaptation

AU - Ham, Yu Jin

AU - Yoo, Chaehwa

AU - Kang, Je Won

N1 - Funding Information: This research was supported by the MSIT(Ministry of Science and ICT), Korea, under the ITRC(Information Technology Research Center) support program(IITP–2021–2020–0–01460) supervised by the IITP(Institute for Information & Communications Technology Planning & Evaluation) Publisher Copyright: © COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.

PY - 2021

Y1 - 2021

N2 - Compression artifact removal is imperative for more visually pleasing contents after image and video compression. Recent works on compression artifact reduction network (CARN) assume that the same or similar quality of images would be employed for both training and testing, and, accordingly, a model needs a quality factor as a prior to accomplish the task successfully. However, the possible discrepancy will degrade performance substantially in a target if the model confronts a different level of distortion from the training phase. To solve the problem, we propose a novel training scheme of CARN to take an advantage of domain adaptation (DA). Specifically, we assign an image encoded with a different quality factor as a different domain and train a CARN using DA to perform robustly in another domain of a different level of distortion. Experimental results demonstrate that the proposed method achieves superior performance on DIV2K, BSD68, and Set12.

AB - Compression artifact removal is imperative for more visually pleasing contents after image and video compression. Recent works on compression artifact reduction network (CARN) assume that the same or similar quality of images would be employed for both training and testing, and, accordingly, a model needs a quality factor as a prior to accomplish the task successfully. However, the possible discrepancy will degrade performance substantially in a target if the model confronts a different level of distortion from the training phase. To solve the problem, we propose a novel training scheme of CARN to take an advantage of domain adaptation (DA). Specifically, we assign an image encoded with a different quality factor as a different domain and train a CARN using DA to perform robustly in another domain of a different level of distortion. Experimental results demonstrate that the proposed method achieves superior performance on DIV2K, BSD68, and Set12.

KW - Compression Artifact Reduction

KW - Deep Learning

KW - Domain Adaptation

KW - Video/Image Coding

UR - http://www.scopus.com/inward/record.url?scp=85118635797&partnerID=8YFLogxK

U2 - 10.1117/12.2597094

DO - 10.1117/12.2597094

M3 - Conference contribution

AN - SCOPUS:85118635797

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Applications of Digital Image Processing XLIV

A2 - Tescher, Andrew G.

A2 - Ebrahimi, Touradj

PB - SPIE

T2 - Applications of Digital Image Processing XLIV 2021

Y2 - 1 August 2021 through 5 August 2021

ER -

Training compression artifacts reduction network with domain adaptation

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Fingerprint

Cite this