SF-CNN: A Fast Compression Artifacts Removal via Spatial-To-Frequency Convolutional Neural Networks

Taeoh Kim; Hyeongmin Lee; Hanbin Son; Sangyoun Lee

doi:10.1109/ICIP.2019.8803503

SF-CNN: A Fast Compression Artifacts Removal via Spatial-To-Frequency Convolutional Neural Networks

Taeoh Kim, Hyeongmin Lee, Hanbin Son, Sangyoun Lee

Department of Electrical and Electronic Engineering

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

14 Citations (Scopus)

Abstract

In this paper, we propose SF-CNN, a fast convolutional neural network structure for JPEG image compression artifacts removal. Recently, Convolutional Neural Network (CNN)-based image restoration has shown great performance improvement. However, its heavy computational cost makes it difficult to apply to other uses such as high-level vision tasks. Since heavy computation arises from maintaining the spatial resolution of an input image, some works make a structure that is composed of spatial downsampling and upsampling operations. SF-CNN takes Spatial input and predicts residual Frequency using downsampling operations only. Since every 8×8 pixel is grouped and spatially invariant in the JPEG DCT domain, it is possible to downsample the input by a factor of 8 to reduce the computational cost. We show this simple structure is effective for compression artifacts removal. Our scalable baseline networks achieve results comparable to to the reference networks in reduced computations.

Original language	English
Title of host publication	2019 IEEE International Conference on Image Processing, ICIP 2019 - Proceedings
Publisher	IEEE Computer Society
Pages	3606-3610
Number of pages	5
ISBN (Electronic)	9781538662496
DOIs	https://doi.org/10.1109/ICIP.2019.8803503
Publication status	Published - 2019 Sept
Event	26th IEEE International Conference on Image Processing, ICIP 2019 - Taipei, Taiwan, Province of China Duration: 2019 Sept 22 → 2019 Sept 25

Publication series

Name	Proceedings - International Conference on Image Processing, ICIP
Volume	2019-September
ISSN (Print)	1522-4880

Conference

Conference	26th IEEE International Conference on Image Processing, ICIP 2019
Country/Territory	Taiwan, Province of China
City	Taipei
Period	19/9/22 → 19/9/25

Bibliographical note

Funding Information:
This research was supported by Multi-Ministry Collaborative R&D Program(R&D program for complex cognitive technology) through the National Research Foundation of Korea(NRF)funded by MSIT, MOTIE,KNPA(NRF-2018M3E3A1057289)

Publisher Copyright:
© 2019 IEEE.

All Science Journal Classification (ASJC) codes

Software
Computer Vision and Pattern Recognition
Signal Processing

Access to Document

10.1109/ICIP.2019.8803503

Cite this

Kim, T., Lee, H., Son, H., & Lee, S. (2019). SF-CNN: A Fast Compression Artifacts Removal via Spatial-To-Frequency Convolutional Neural Networks. In 2019 IEEE International Conference on Image Processing, ICIP 2019 - Proceedings (pp. 3606-3610). [8803503] (Proceedings - International Conference on Image Processing, ICIP; Vol. 2019-September). IEEE Computer Society. https://doi.org/10.1109/ICIP.2019.8803503

@inproceedings{e2942cd334ed43dc8add344188da9c9e,

title = "SF-CNN: A Fast Compression Artifacts Removal via Spatial-To-Frequency Convolutional Neural Networks",

abstract = "In this paper, we propose SF-CNN, a fast convolutional neural network structure for JPEG image compression artifacts removal. Recently, Convolutional Neural Network (CNN)-based image restoration has shown great performance improvement. However, its heavy computational cost makes it difficult to apply to other uses such as high-level vision tasks. Since heavy computation arises from maintaining the spatial resolution of an input image, some works make a structure that is composed of spatial downsampling and upsampling operations. SF-CNN takes Spatial input and predicts residual Frequency using downsampling operations only. Since every 8×8 pixel is grouped and spatially invariant in the JPEG DCT domain, it is possible to downsample the input by a factor of 8 to reduce the computational cost. We show this simple structure is effective for compression artifacts removal. Our scalable baseline networks achieve results comparable to to the reference networks in reduced computations.",

author = "Taeoh Kim and Hyeongmin Lee and Hanbin Son and Sangyoun Lee",

note = "Funding Information: This research was supported by Multi-Ministry Collaborative R&D Program(R&D program for complex cognitive technology) through the National Research Foundation of Korea(NRF)funded by MSIT, MOTIE,KNPA(NRF-2018M3E3A1057289) Publisher Copyright: {\textcopyright} 2019 IEEE.; 26th IEEE International Conference on Image Processing, ICIP 2019 ; Conference date: 22-09-2019 Through 25-09-2019",

year = "2019",

month = sep,

doi = "10.1109/ICIP.2019.8803503",

language = "English",

series = "Proceedings - International Conference on Image Processing, ICIP",

publisher = "IEEE Computer Society",

pages = "3606--3610",

booktitle = "2019 IEEE International Conference on Image Processing, ICIP 2019 - Proceedings",

address = "United States",

}

Kim, T, Lee, H, Son, H & Lee, S 2019, SF-CNN: A Fast Compression Artifacts Removal via Spatial-To-Frequency Convolutional Neural Networks. in 2019 IEEE International Conference on Image Processing, ICIP 2019 - Proceedings., 8803503, Proceedings - International Conference on Image Processing, ICIP, vol. 2019-September, IEEE Computer Society, pp. 3606-3610, 26th IEEE International Conference on Image Processing, ICIP 2019, Taipei, Taiwan, Province of China, 19/9/22. https://doi.org/10.1109/ICIP.2019.8803503

SF-CNN : A Fast Compression Artifacts Removal via Spatial-To-Frequency Convolutional Neural Networks. / Kim, Taeoh; Lee, Hyeongmin; Son, Hanbin et al.

2019 IEEE International Conference on Image Processing, ICIP 2019 - Proceedings. IEEE Computer Society, 2019. p. 3606-3610 8803503 (Proceedings - International Conference on Image Processing, ICIP; Vol. 2019-September).

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

TY - GEN

T1 - SF-CNN

T2 - 26th IEEE International Conference on Image Processing, ICIP 2019

AU - Kim, Taeoh

AU - Lee, Hyeongmin

AU - Son, Hanbin

AU - Lee, Sangyoun

N1 - Funding Information: This research was supported by Multi-Ministry Collaborative R&D Program(R&D program for complex cognitive technology) through the National Research Foundation of Korea(NRF)funded by MSIT, MOTIE,KNPA(NRF-2018M3E3A1057289) Publisher Copyright: © 2019 IEEE.

PY - 2019/9

Y1 - 2019/9

N2 - In this paper, we propose SF-CNN, a fast convolutional neural network structure for JPEG image compression artifacts removal. Recently, Convolutional Neural Network (CNN)-based image restoration has shown great performance improvement. However, its heavy computational cost makes it difficult to apply to other uses such as high-level vision tasks. Since heavy computation arises from maintaining the spatial resolution of an input image, some works make a structure that is composed of spatial downsampling and upsampling operations. SF-CNN takes Spatial input and predicts residual Frequency using downsampling operations only. Since every 8×8 pixel is grouped and spatially invariant in the JPEG DCT domain, it is possible to downsample the input by a factor of 8 to reduce the computational cost. We show this simple structure is effective for compression artifacts removal. Our scalable baseline networks achieve results comparable to to the reference networks in reduced computations.

AB - In this paper, we propose SF-CNN, a fast convolutional neural network structure for JPEG image compression artifacts removal. Recently, Convolutional Neural Network (CNN)-based image restoration has shown great performance improvement. However, its heavy computational cost makes it difficult to apply to other uses such as high-level vision tasks. Since heavy computation arises from maintaining the spatial resolution of an input image, some works make a structure that is composed of spatial downsampling and upsampling operations. SF-CNN takes Spatial input and predicts residual Frequency using downsampling operations only. Since every 8×8 pixel is grouped and spatially invariant in the JPEG DCT domain, it is possible to downsample the input by a factor of 8 to reduce the computational cost. We show this simple structure is effective for compression artifacts removal. Our scalable baseline networks achieve results comparable to to the reference networks in reduced computations.

UR - http://www.scopus.com/inward/record.url?scp=85076809484&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85076809484&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2019.8803503

DO - 10.1109/ICIP.2019.8803503

M3 - Conference contribution

AN - SCOPUS:85076809484

T3 - Proceedings - International Conference on Image Processing, ICIP

SP - 3606

EP - 3610

BT - 2019 IEEE International Conference on Image Processing, ICIP 2019 - Proceedings

PB - IEEE Computer Society

Y2 - 22 September 2019 through 25 September 2019

ER -

SF-CNN: A Fast Compression Artifacts Removal via Spatial-To-Frequency Convolutional Neural Networks

Abstract

Publication series

Conference

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this