Discussion of a Pruning Scheme for Top-K Retrievals Among Vector Quantizer Encoded Signatures

Anindya Sarkar (1), Vishwakarma Singh (2), Pratim Ghosh (1), B. S. Manjunath (1), Ambuj Singh(2)
(1) Department of Electrical and Computer Engineering, University of California, Santa Barbara
(2) Department of Computer Science, University of California, Santa Barbara


The problem we are considering here is duplicate video detection. We have a database of N videos and we store compact signatures, called fingerprints, for each of them. When a query video is presented, the system first returns the top-K most closely matched videos. Then, a more detailed search is performed among the top-K retrieved model videos to obtain the best match. Finally, a separate module is used to confirm whether the best matched video is indeed a duplicate. In this write-up, we focus on the VQ based pruned search where the effort is to return the top-K neighbors in the fastest possible manner, without having to do a linear scan of all the N database signatures.
[PDF] [BibTex]
Anindya Sarkar, Vishwakarma Singh, Pratim Ghosh, B. S. Manjunath and Ambuj Singh,
VRL, ECE, University of California, Santa Barbara, Apr. 2008.
Node ID: 494 , DB ID: 301 , Lab: VRL , Target: Technical Report
Subject: [Managing Multimedia Databases] « Look up more