Search | Introduction | Publications | VIP Lab | UC Berkeley

Video Cluster Search

Search Terms (case insensitive):

A note about the drawing : this drawing illustrates the basic principle behind Video Signature, the technique used in this search engine to summarize each video. The v's and w's represent video frames from two visually similar video sequences. The video signature for each video consists of the frames that are closest to a small set of random vectors such as s1 and s2. It can be shown that by comparing the corresponding frames in two video signatures, the fraction of similar frames between two video sequences can be reliably measured.
Samson Cheung
Copyright 1999-2003. All rights reserved.
Last modified: Sat Dec 21 12:25:19 PST 2002