A new similarity measure for near duplicate video clip detection

  • Authors:

  • Xiangmin Zhou;Xiaofang Zhou;Heng Tao Shen

  • Affiliations:

  • School of Information Technology & Electrical Engineering, University of Queensland, Australia;School of Information Technology & Electrical Engineering, University of Queensland, Australia;School of Information Technology & Electrical Engineering, University of Queensland, Australia

  • Venue:
  • APWeb/WAIM'07 Proceedings of the joint 9th Asia-Pacific web and 8th international conference on web-age information management conference on Advances in data and web management
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Near-duplicate video clip(NDVC) detection is a special issue of content-based video search. Identifying the videos derived from the same original source is the primary task of this research. In NDVC detection, an important step is to define an effective similarity measure that captures both frame and sequence information inherent to the video clips. To address this, in this paper, we propose a new similarity measure, named as Video Edit Distance(VED), that adopts a complementary information compensation scheme based on the visual features and sequence context of videos. Visual features contain the discriminative information of each video, and sequence context captures the feature variation of it. To reduce the computation cost of inter-video comparison by VED, we extract key frames from video sequences and map each key frame into one single symbol. Various techniques are proposed to compensate the information loss in the measurement. Experimental results demonstrate that the proposed measure is highly effective.