VIDCAR: An Improved CBVR System Exploiting the Dynamics in MST-CSS Feature Descriptor

Chiranjoy Chattopadhyay and Sukhendu Das
Visualization and Perception Lab
Department of Computer Science and Engineering, Indian Institute of Technology, Madras, India

Accepted with minor correction in International Journal of Multimedia Information Retrieval (IJMIR), Springer, August 2014

Abstract

This paper presents VIDCAR (VIDeo Content Analysis and Retrieval), an unsupervised framework for Content Based Video Retrieval (CBVR) using representation of the dynamics in the spatio-temporal model extracted from video shots. We propose DMST-CSS (Dynamic Multi Spectro Temporal- Curvature Scale Space), an improved feature descriptor for enhancing the performance of CBVR task. Our primary contribution is in representation of the dynamics of the evolution of the MST-CSS surface. Unlike the earlier MST-CSS descriptor, which extracts geometric features after the evolving MST-CSS surface converges to a final formation, this DMST-CSS captures the dynamics of the evolution (formation) of the surface and is thus more robust. We have represented the evolution (or dynamics) of MST-CSS surface as a multivariate time series to obtain a DMST-CSS descriptor. A global kernel alignment technique has been adapted to compute a match cost between query and model DMST-CSS descriptor. Performance comparison with existing CBVR techniques on five benchmark video datasets proves the superiority of VIDCAR.

Framework


VICAR framwork

Evolution of MST-CSS Surface


Animation of the evolution of the hilly MST-CSS surface over time.

Clck the links below to view and download the videos (in a new page)

bullet Set 1

Set 2

Set 3