DeepSIC: Deep semantic image compression

Sihui Luo; Yezhou Yang; Yanling Yin; Chengchao Shen; Ya Zhao; Mingli Song

doi:10.1007/978-3-030-04167-0_9

DeepSIC: Deep semantic image compression

Sihui Luo, Yezhou Yang, Yanling Yin, Chengchao Shen, Ya Zhao, Mingli Song

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

25 Scopus citations

Abstract

Incorporating semantic analysis into image compression can significantly reduce the repetitive computation of fundamental semantic analysis in client-side applications such as semantic image retrieval. The same practice also enables the compressed code to carry semantic information of the image during its storage and transmission. In this paper, we propose a Deep Semantic Image Compression (DeepSIC) model to achieve this goal and put forward two novel architectures that aim to reconstruct the compressed image and generate corresponding semantic representations at the same time by a single end-to-end optimized network. The first architecture performs semantic analysis in the encoding process by reserving a portion of the bits from the compressed code to store the semantic representations. The second performs semantic analysis in the decoding step with the feature maps that are embedded in the compressed code. In both architectures, the feature maps are shared by the compression and the semantic analytics modules. Experiments over benchmarking datasets show promising performance of the proposed compression model.

Original language	English (US)
Title of host publication	Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings
Editors	Long Cheng, Andrew Chi Sing Leung, Seiichi Ozawa
Publisher	Springer Verlag
Pages	96-106
Number of pages	11
ISBN (Print)	9783030041663
DOIs	https://doi.org/10.1007/978-3-030-04167-0_9
State	Published - 2018
Event	25th International Conference on Neural Information Processing, ICONIP 2018 - Siem Reap, Cambodia Duration: Dec 13 2018 → Dec 16 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11301 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	25th International Conference on Neural Information Processing, ICONIP 2018
Country/Territory	Cambodia
City	Siem Reap
Period	12/13/18 → 12/16/18

Keywords

Deep image compression
End-to-end optimization
Semantic image compression

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-030-04167-0_9

Cite this

Luo, S., Yang, Y., Yin, Y., Shen, C., Zhao, Y., & Song, M. (2018). DeepSIC: Deep semantic image compression. In L. Cheng, A. C. S. Leung, & S. Ozawa (Eds.), Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings (pp. 96-106). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11301 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-04167-0_9

DeepSIC: Deep semantic image compression. / Luo, Sihui; Yang, Yezhou; Yin, Yanling et al.
Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. ed. / Long Cheng; Andrew Chi Sing Leung; Seiichi Ozawa. Springer Verlag, 2018. p. 96-106 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11301 LNCS).

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

Luo, S, Yang, Y, Yin, Y, Shen, C, Zhao, Y & Song, M 2018, DeepSIC: Deep semantic image compression. in L Cheng, ACS Leung & S Ozawa (eds), Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11301 LNCS, Springer Verlag, pp. 96-106, 25th International Conference on Neural Information Processing, ICONIP 2018, Siem Reap, Cambodia, 12/13/18. https://doi.org/10.1007/978-3-030-04167-0_9

Luo S, Yang Y, Yin Y, Shen C, Zhao Y, Song M. DeepSIC: Deep semantic image compression. In Cheng L, Leung ACS, Ozawa S, editors, Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. Springer Verlag. 2018. p. 96-106. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-04167-0_9

Luo, Sihui ; Yang, Yezhou ; Yin, Yanling et al. / DeepSIC : Deep semantic image compression. Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. editor / Long Cheng ; Andrew Chi Sing Leung ; Seiichi Ozawa. Springer Verlag, 2018. pp. 96-106 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{fea6f71050004c8c8c495cb4ad9280e7,

title = "DeepSIC: Deep semantic image compression",

abstract = "Incorporating semantic analysis into image compression can significantly reduce the repetitive computation of fundamental semantic analysis in client-side applications such as semantic image retrieval. The same practice also enables the compressed code to carry semantic information of the image during its storage and transmission. In this paper, we propose a Deep Semantic Image Compression (DeepSIC) model to achieve this goal and put forward two novel architectures that aim to reconstruct the compressed image and generate corresponding semantic representations at the same time by a single end-to-end optimized network. The first architecture performs semantic analysis in the encoding process by reserving a portion of the bits from the compressed code to store the semantic representations. The second performs semantic analysis in the decoding step with the feature maps that are embedded in the compressed code. In both architectures, the feature maps are shared by the compression and the semantic analytics modules. Experiments over benchmarking datasets show promising performance of the proposed compression model.",

keywords = "Deep image compression, End-to-end optimization, Semantic image compression",

author = "Sihui Luo and Yezhou Yang and Yanling Yin and Chengchao Shen and Ya Zhao and Mingli Song",

note = "Funding Information: Acknowledgment. This work is supported by National Natural Science Foundation of China (61572428, U1509206), Fundamental Research Funds for the Central Universities (2017FZA5014), National Key Research and Development Program (2016YFB1200203) and Key Research and Development Program of Zhejiang Province (2018C01004).; 25th International Conference on Neural Information Processing, ICONIP 2018 ; Conference date: 13-12-2018 Through 16-12-2018",

year = "2018",

doi = "10.1007/978-3-030-04167-0_9",

language = "English (US)",

isbn = "9783030041663",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "96--106",

editor = "Long Cheng and Leung, {Andrew Chi Sing} and Seiichi Ozawa",

booktitle = "Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings",

}

TY - GEN

T1 - DeepSIC

T2 - 25th International Conference on Neural Information Processing, ICONIP 2018

AU - Luo, Sihui

AU - Yang, Yezhou

AU - Yin, Yanling

AU - Shen, Chengchao

AU - Zhao, Ya

AU - Song, Mingli

N1 - Funding Information: Acknowledgment. This work is supported by National Natural Science Foundation of China (61572428, U1509206), Fundamental Research Funds for the Central Universities (2017FZA5014), National Key Research and Development Program (2016YFB1200203) and Key Research and Development Program of Zhejiang Province (2018C01004).

PY - 2018

Y1 - 2018

N2 - Incorporating semantic analysis into image compression can significantly reduce the repetitive computation of fundamental semantic analysis in client-side applications such as semantic image retrieval. The same practice also enables the compressed code to carry semantic information of the image during its storage and transmission. In this paper, we propose a Deep Semantic Image Compression (DeepSIC) model to achieve this goal and put forward two novel architectures that aim to reconstruct the compressed image and generate corresponding semantic representations at the same time by a single end-to-end optimized network. The first architecture performs semantic analysis in the encoding process by reserving a portion of the bits from the compressed code to store the semantic representations. The second performs semantic analysis in the decoding step with the feature maps that are embedded in the compressed code. In both architectures, the feature maps are shared by the compression and the semantic analytics modules. Experiments over benchmarking datasets show promising performance of the proposed compression model.

AB - Incorporating semantic analysis into image compression can significantly reduce the repetitive computation of fundamental semantic analysis in client-side applications such as semantic image retrieval. The same practice also enables the compressed code to carry semantic information of the image during its storage and transmission. In this paper, we propose a Deep Semantic Image Compression (DeepSIC) model to achieve this goal and put forward two novel architectures that aim to reconstruct the compressed image and generate corresponding semantic representations at the same time by a single end-to-end optimized network. The first architecture performs semantic analysis in the encoding process by reserving a portion of the bits from the compressed code to store the semantic representations. The second performs semantic analysis in the decoding step with the feature maps that are embedded in the compressed code. In both architectures, the feature maps are shared by the compression and the semantic analytics modules. Experiments over benchmarking datasets show promising performance of the proposed compression model.

KW - Deep image compression

KW - End-to-end optimization

KW - Semantic image compression

UR - http://www.scopus.com/inward/record.url?scp=85059074383&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059074383&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-04167-0_9

DO - 10.1007/978-3-030-04167-0_9

M3 - Conference contribution

AN - SCOPUS:85059074383

SN - 9783030041663

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 96

EP - 106

BT - Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings

A2 - Cheng, Long

A2 - Leung, Andrew Chi Sing

A2 - Ozawa, Seiichi

PB - Springer Verlag

Y2 - 13 December 2018 through 16 December 2018

ER -

DeepSIC: Deep semantic image compression

Abstract

Publication series

Other

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this