Skip to Content
Find More Like This
Return to Search

Using dynamic mode decomposition for real-time background/foreground separation in video

United States Patent

June 6, 2017
View the Complete Patent at the US Patent & Trademark Office
The technique of dynamic mode decomposition (DMD) is disclosed herein for the purpose of robustly separating video frames into background (low-rank) and foreground (sparse) components in real-time. Foreground/background separation is achieved at the computational cost of just one singular value decomposition (SVD) and one linear equation solve, thus producing results orders of magnitude faster than robust principal component analysis (RPCA). Additional techniques, including techniques for analyzing the video for multi-resolution time-scale components, and techniques for reusing computations to allow processing of streaming video in real time, are also described herein.
Kutz; Jose Nathan (Seattle, WA), Grosek; Jacob (Seattle, WA), Brunton; Steven (Seattle, WA), Fu; Xing (Seattle, WA), Pendergrass; Seth (Seattle, WA)
14/ 828,396
August 17, 2015
STATEMENT OF GOVERNMENT LICENSE RIGHTS This invention was made with Government support under Grant No. DE-EE0006785, awarded by the Department of Energy (DOE); Grant No. FA9550-09-0174, awarded by the Air Force Office of Scientific Research (AFOSR); and Grant No. DMS-100762, awarded by the National Science Foundation (NSF). The Government has certain rights in the invention.