An Object-Based Approach for Digital Video Retrieval

  • Authors:
  • Mark Smith;Alireza Khotanzad

  • Affiliations:
  • -;-

  • Venue:
  • ITCC '04 Proceedings of the International Conference on Information Technology: Coding and Computing (ITCC'04) Volume 2 - Volume 2
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

A new method for retrieving video segments from adigital video database is proposed. A unique signature foreach video segment is generated from the analysis andtracking of interesting objects between each frame of thevideo region. The MPEG-7 descriptors [Introduction to MPEG-7] consisting ofthe edge histogram, homogenous texture, dominant colorsand color structure (188 features in all) are extractedfrom corresponding objects in successive frames for theduration of the considered sequence. The average of theMPEG-7 descriptors is computed for each object acrossall frames in the region thus resulting in a uniquesignature for the region. The stored signature which bestmatches this new signature (based on the earth moverdistance [The Earth's Mover's Distance, Multi-Dimensional Scaling, and Color-Based Image Retrieval] between the signatures) is used for retrievingthe video segment.