Title: A content-based video retrieval system: video retrieval with extensive features

Authors: Priya Rajendran, T.N. Shanmugam

Addresses: Department of Mathematics, Anna University – Chennai, Chennai – 600 025, Tamilnadu, India. ' Department of Mathematics, Anna University – Chennai, Chennai – 600 025, Tamilnadu, India

Abstract: The existing video retrieval techniques are inefficient because they utilise a specific feature. In this paper, a proficient system with extensive features is proposed for improving the retrieval efficiency. The system segments a video into shots, and then a few representative frames are extracted from each shot, and colour, contour, texture and motion features-based frame descriptors are computed for these shots and stored in a feature library. When a query clip is given, the aforesaid features are extracted for the query clip and compared with the features that are already stored in the feature library. Latent semantic indexing (LSI)-based similarity measure is used for performing the comparison. Finally, videos that are similar to the query are retrieved from the enormous collection of videos. The proposed system is evaluated with precision-recall and F score and they are compared with existing retrieval systems.

Keywords: content-based video retrieval; CBVR; shot segmentation; motion estimation; Gaussian mixture model; GMM; colour histogram; query clips; latent semantic indexing; LSI; multimedia; extensive features; feature libraries; similarity measures.

DOI: 10.1504/IJMIS.2011.041363

International Journal of Multimedia Intelligence and Security, 2011 Vol.2 No.2, pp.146 - 171

Published online: 28 Feb 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article