Content-based Video Copy Detection using Binary Object Fingerprints

Mengyang Liu, Lai-Man Po, Chang Zhou, Wilson Y. F. Yuen, Hok-Kwan Cheung, Peter H. W. Wong, Hon-Tung Luk, Kin-Wai Lau

Research output: Chapters, Conference Papers, Creative and Literary WorksRGC 32 - Refereed conference paper (with host publication)peer-review

5 Citations (Scopus)

Abstract

Content-based video copy detection in large-scale databases is still an open issue. One of the main reasons is that geometrical attack can easily surpass the global features of the frame while local features are inefficient in terms of compact representation and computational complexity. In this paper, we propose to use binary object fingerprints to represent video frame for improving the robustness of the video copy detection system. It is because salient object can be robustly detected using advanced convolutional neural network (CNN) based object detector. We proposed to use the well-known RetinaNet for generating object regions from the input frame and then these regions are used to generate binary fingerprints for fast copy detection in the database. This approach can maintain compact representation of video frame and high searching speed by binary fingerprint searching scheme. Experimental results show that the proposed approach can achieve about 10% higher recall rate with only sacrificing 1% prediction rate on VCDB dataset.
Original languageEnglish
Title of host publication2018 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC 2018) - Conference Proceedings
PublisherIEEE
ISBN (Electronic)978-1-5386-7946-3
DOIs
Publication statusPublished - Sept 2018
Event8th IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2018 - Qingdao, China
Duration: 14 Sept 201816 Sept 2018

Publication series

NameIEEE International Conference on Signal Processing, Communications and Computing, ICSPCC

Conference

Conference8th IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2018
PlaceChina
CityQingdao
Period14/09/1816/09/18

Research Keywords

  • CNN
  • Content-based Copy Detection
  • Object Detection
  • Video Fingerprint

Fingerprint

Dive into the research topics of 'Content-based Video Copy Detection using Binary Object Fingerprints'. Together they form a unique fingerprint.

Cite this