Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation

Manar Abduljabbar Ahmad Mizher, Mei Choo Ang, Siti Norul Huda Sheikh Abdullah, Kok Weng Ng

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

Key frame extraction is an important algorithm for video summarisation, video retrieval, and generating video fingerprint. The extracted key frames should represent a video sequence in a compact way and brief the main actions to achieve meaningful key frames. Therefore, we present a key frames extraction algorithm based on the L1-norm by accumulating action frames via optical flow method. We then evaluate our proposed algorithm using the action accuracy rate and action error rate of the extracted action frames in comparison to user extraction. The video shot summarisation evaluation shows that our proposed algorithm outperforms the-state-of-the-art algorithms in terms of compression ratio. Our proposed algorithm also achieves approximately 100% and 0.91% for best and worst case in terms of action appearance accuracy in human action dataset KTH in the extracted key frames.

Original languageEnglish
Title of host publicationAdvances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings
PublisherSpringer Verlag
Pages364-375
Number of pages12
Volume10645 LNCS
ISBN (Print)9783319700090
DOIs
Publication statusPublished - 1 Jan 2017
Event5th International Visual Informatics Conference, IVIC 2017 - Bangi, Malaysia
Duration: 28 Nov 201730 Nov 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10645 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other5th International Visual Informatics Conference, IVIC 2017
CountryMalaysia
CityBangi
Period28/11/1730/11/17

Fingerprint

L1-norm
Optical flows
Optical Flow
Summarization
Video Summarization
Video Retrieval
Fingerprint
Error Rate
Compression
Evaluate
Evaluation

Keywords

  • Blocks differential
  • Colour histogram
  • Frame differences
  • L1-norm
  • Optical flow

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Mizher, M. A. A., Ang, M. C., Sheikh Abdullah, S. N. H., & Ng, K. W. (2017). Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation. In Advances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings (Vol. 10645 LNCS, pp. 364-375). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10645 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-70010-6_34

Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation. / Mizher, Manar Abduljabbar Ahmad; Ang, Mei Choo; Sheikh Abdullah, Siti Norul Huda; Ng, Kok Weng.

Advances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings. Vol. 10645 LNCS Springer Verlag, 2017. p. 364-375 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10645 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Mizher, MAA, Ang, MC, Sheikh Abdullah, SNH & Ng, KW 2017, Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation. in Advances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings. vol. 10645 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10645 LNCS, Springer Verlag, pp. 364-375, 5th International Visual Informatics Conference, IVIC 2017, Bangi, Malaysia, 28/11/17. https://doi.org/10.1007/978-3-319-70010-6_34
Mizher MAA, Ang MC, Sheikh Abdullah SNH, Ng KW. Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation. In Advances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings. Vol. 10645 LNCS. Springer Verlag. 2017. p. 364-375. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-70010-6_34
Mizher, Manar Abduljabbar Ahmad ; Ang, Mei Choo ; Sheikh Abdullah, Siti Norul Huda ; Ng, Kok Weng. / Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation. Advances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings. Vol. 10645 LNCS Springer Verlag, 2017. pp. 364-375 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{0391847d03b847fb97f0b9d41a6873ac,
title = "Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation",
abstract = "Key frame extraction is an important algorithm for video summarisation, video retrieval, and generating video fingerprint. The extracted key frames should represent a video sequence in a compact way and brief the main actions to achieve meaningful key frames. Therefore, we present a key frames extraction algorithm based on the L1-norm by accumulating action frames via optical flow method. We then evaluate our proposed algorithm using the action accuracy rate and action error rate of the extracted action frames in comparison to user extraction. The video shot summarisation evaluation shows that our proposed algorithm outperforms the-state-of-the-art algorithms in terms of compression ratio. Our proposed algorithm also achieves approximately 100{\%} and 0.91{\%} for best and worst case in terms of action appearance accuracy in human action dataset KTH in the extracted key frames.",
keywords = "Blocks differential, Colour histogram, Frame differences, L1-norm, Optical flow",
author = "Mizher, {Manar Abduljabbar Ahmad} and Ang, {Mei Choo} and {Sheikh Abdullah}, {Siti Norul Huda} and Ng, {Kok Weng}",
year = "2017",
month = "1",
day = "1",
doi = "10.1007/978-3-319-70010-6_34",
language = "English",
isbn = "9783319700090",
volume = "10645 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "364--375",
booktitle = "Advances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings",
address = "Germany",

}

TY - GEN

T1 - Action key frames extraction using L1-Norm and accumulative optical flow for compact video shot summarisation

AU - Mizher, Manar Abduljabbar Ahmad

AU - Ang, Mei Choo

AU - Sheikh Abdullah, Siti Norul Huda

AU - Ng, Kok Weng

PY - 2017/1/1

Y1 - 2017/1/1

N2 - Key frame extraction is an important algorithm for video summarisation, video retrieval, and generating video fingerprint. The extracted key frames should represent a video sequence in a compact way and brief the main actions to achieve meaningful key frames. Therefore, we present a key frames extraction algorithm based on the L1-norm by accumulating action frames via optical flow method. We then evaluate our proposed algorithm using the action accuracy rate and action error rate of the extracted action frames in comparison to user extraction. The video shot summarisation evaluation shows that our proposed algorithm outperforms the-state-of-the-art algorithms in terms of compression ratio. Our proposed algorithm also achieves approximately 100% and 0.91% for best and worst case in terms of action appearance accuracy in human action dataset KTH in the extracted key frames.

AB - Key frame extraction is an important algorithm for video summarisation, video retrieval, and generating video fingerprint. The extracted key frames should represent a video sequence in a compact way and brief the main actions to achieve meaningful key frames. Therefore, we present a key frames extraction algorithm based on the L1-norm by accumulating action frames via optical flow method. We then evaluate our proposed algorithm using the action accuracy rate and action error rate of the extracted action frames in comparison to user extraction. The video shot summarisation evaluation shows that our proposed algorithm outperforms the-state-of-the-art algorithms in terms of compression ratio. Our proposed algorithm also achieves approximately 100% and 0.91% for best and worst case in terms of action appearance accuracy in human action dataset KTH in the extracted key frames.

KW - Blocks differential

KW - Colour histogram

KW - Frame differences

KW - L1-norm

KW - Optical flow

UR - http://www.scopus.com/inward/record.url?scp=85035136944&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85035136944&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-70010-6_34

DO - 10.1007/978-3-319-70010-6_34

M3 - Conference contribution

SN - 9783319700090

VL - 10645 LNCS

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

SP - 364

EP - 375

BT - Advances in Visual Informatics - 5th International Visual Informatics Conference, IVIC 2017, Proceedings

PB - Springer Verlag

ER -